1 s2.0 S0022509618307439 Main

Journal of the Mechanics and Physics of Solids 125 (2019) 298–325
Contents lists available at ScienceDirect
Journal of the Mechanics and Physics of Solids

journal homepage: www.elsevier.com/locate/jmps
Mechanobiological stability of biological soft tissues

Marcos Latorre a,∗, Jay D. Humphrey a,b
a
Department of Biomedical Engineering, Yale University, New Haven, CT 06520, USA
b
Vascular Biology and Therapeutics Program, Yale School of Medicine, New Haven, CT 06520, USA
a r t i c l e i n f o a b s t r a c t
Article history: Like all other materials, biological soft tissues are subject to general laws of physics, in-
Received 30 August 2018 cluding those governing mechanical equilibrium and stability. In addition, however, these
Revised 28 November 2018
tissues are able to respond actively to changes in their mechanical and chemical envi-
Accepted 20 December 2018
ronment. There is, therefore, a pressing need to understand such processes theoretically.
Available online 21 December 2018
In this paper, we present a new rate-based constrained mixture formulation suitable for
Keywords: studying mechanobiological equilibrium and stability of soft tissues exposed to transient
Mechanical homeostasis or sustained changes in material composition or applied loading. These concepts are il-
Extracellular matrix lustrated for canonical problems in arterial mechanics, which distinguish possible stable
Adaptation versus unstable mechanobiological responses. Such analyses promise to yield insight into
Matrix turnover biological processes that govern both health and disease progression.
Tissue growth
© 2018 Elsevier Ltd. All rights reserved.
1. Introduction
Biological soft tissues consist of myriad proteins, glycoproteins, and glycosaminoglycans (Hynes and Naba 2012), each
having individual natural (stress-free) configurations, mechanical properties, and rates of turnover. It is appropriate, there-
fore, to consider mixture theories of multiple solid constituents when establishing theoretical frameworks for mathemati-
cally modeling growth (changes in mass) and remodeling (changes in microstructure). Adopting a classical continuum the-
ory of mixtures presents numerous challenges, however, including difficulty in identifying constitutive relations for linear
momentum exchanges between constituents as they turnover and challenges in prescribing how traction boundary condi-
tions partition by constituent, particularly as they evolve. For this reason, we have advocated a constrained mixture theory
wherein one assumes that structurally significant constituents possess individual natural configurations but are constrained
to move with the mixture as a whole, and one satisfies full mixture equations for mass balance but classical equations
for linear momentum balance augmented with a rule-of-mixture relation for the stored energy (Humphrey and Rajagopal
2002). Such models simplify the constitutive formulation and solution of initial-boundary value problems. Nevertheless, the
full constrained mixture theory involves heredity integrals for the evolution of constituent-specific mass density and stress,
which can render the associated computational modeling expensive except in problems defined by simple geometries. Rate-
based models can thus be useful.
It is also well-known that biological soft tissues grow and remodel in response to changes in mechanical loading. Such
responses – that is, changes in microstructural composition and/or organization and thus changes in mechanical proper-
ties and geometry – stem from mechanobiological processes, often changes in gene expression that control the production
and removal of constituents in potentially evolving configurations (Humphrey et al. 2014). In parallel to classical concepts
∗
Corresponding author.
E-mail addresses: marcos.latorre@yale.edu (M. Latorre), jay.humphrey@yale.edu (J.D. Humphrey).
https://doi.org/10.1016/j.jmps.2018.12.013
0022-5096/© 2018 Elsevier Ltd. All rights reserved.
M. Latorre and J.D. Humphrey / Journal of the Mechanics and Physics of Solids 125 (2019) 298–325 299
of mechanical equilibrium and mechanical stability, modeling biological growth and remodeling (G&R) necessitates an un-
derstanding of mechanobiological equilibrium and mechanobiological stability. The former can be defined by a balanced
production and removal of stressed constituents within an unchanging configuration (Latorre and Humphrey 2018b); the
latter can be defined as the ability of a tissue to preserve its mechanobiological equilibrium despite transient perturbations
in loading under physiological or pathophysiological conditions (Cyron and Humphrey 2014). For example, we do not expect
a muscle to grow or remodel simply because of a transient loading. Rather, we expect a muscle to grow or remodel in
response to sustained or repetitive loading, which could lead to mechanobiological adaptivity to the new mechanical en-
vironment. Failed adaptivity can occur in disease and injury, however, which may result from a loss of mechanobiological
stability.
In this paper, we first summarize a full hereditary integral-based constrained mixture model for growth and remodeling
of soft tissues. We then derive a fully three-dimensional rate-based constrained mixture theory that is equivalent to the
integral-based model. We subsequently particularize this formulation to a prototypical cylindrical artery to identify equilib-
rium solutions for vanishing rates and to assess their mechanobiological (static) stability with respect to sustained changes
in external loads or model parameters. Finally, we use this rate-based approach to assess whether a (dynamic, self-excited)
G&R process around a previously equilibrated solution is mechanobiologically unstable or stable, either neutrally or asymp-
totically, with respect to perturbations in external loads that are eventually sustained over time. For purposes of illustration,
we consider computational results for a cylindrical murine aorta exposed to sustained or transient changes in mechanical
loading and/or material properties. Consistent with most prior constitutive descriptions of arterial mechanics, we assume
an underlying pseudoelastic rather than viscoelastic or poroelastic material response; similarly, consistent with most prior
stress analyses, we assume quasi-static solutions rather than elastodynamics over a cardiac cycle (Humphrey 2002). In this
way, we focus primarily on evolving mechanobiologically induced changes in geometry and material properties. Finally, we
do not consider further complications associated with aortic dissection or rupture, which necessarily require one to invoke
additional constitutive frameworks.
2. Theoretical framework
2.1. Constrained mixture framework: Integral form
Consider the Cauchy stress σ , at G&R time s, for a biological soft tissue consisting of a mixture of N structurally significant
constituents and exhibiting incompressible behavior under transient mechanical loading,

N
σ(s ) = σ α(s ) − p(s )I (1)
α =1
with σ α the mixture-level contribution to σ of constituent α = 1, . . . , N and p the Lagrange multiplier that enforces tran-
sient incompressibility although the tissue may change mass and volume over G&R timescales. The stress tensor σ α can
be expressed in constrained mixture models (Humphrey and Rajagopal 2002) through the following hereditary integral
(Latorre and Humphrey 2018b)
s
1 α
σ α(s ) = mα(τ )qα(s, τ )σˆ (s, τ )dτ (2)
ρ
−∞
where ρ is the assumed constant spatial mass density of the tissue (since most of the structurally significant constituents
are hydrated similarly), mα (τ ) > 0 is a spatial mass density production rate of constituent α , qα(s, τ ) ∈ [0, 1] is a survival
α
function for constituent α deposited within extant matrix at G&R time τ ≤ s that survives at G&R time s, and σˆ (s, τ ) is
ˆ α (Cα (s ) ) be the strain energy function of
the constituent-level Cauchy stress at s for constituent α deposited at τ . Let W n (τ )
constituent α , with Cα s = Fα
n (τ )( )
T s Fα s the right Cauchy–Green tensor at time s for constituent α deposited at time τ ,
n (τ )( ) n (τ )( )
which is computed from the associated deformation gradient Fα s with respect to natural configuration κnα (τ ), which is
n (τ )( )
denoted n(τ ). It can be shown that (Baek et al. 2006)
Fαn(τ )(s ) = F(s )F−1(τ )Gα(τ ) (3)
with F the deformation gradient of the mixture at time s or τ , measured with respect to an original homeostatic (reference)
configuration κ (0) (Fig. 1), and Gα (τ ) a symmetric (Gα = Gα T ) and volume-preserving (det Gα = 1) deposition (pre)stretch
tensor by which constituent α is deposited within the extant matrix at time τ relative to its own possibly evolving natural
configuration κnα(τ ) (Fig. 1). This deposition stretch arises via synthetic cells acting on the secreted matrix via actomyosin
activity (Humphrey et al. 2014), thus its magnitude becomes a constitutive parameter and so too the orientation of the new
tissue when deposited (Baek et al. 2006; Valentín et al. 2013). Furthermore, let the constituent-level Cauchy stress for each
constituent α be
2 ∂ Wˆ α (Cαn(τ )(s ) ) αT
σˆ α(s, τ ) = Fα ( s ) Fn(τ )(s ) (4)
Jnα(τ )(s ) n(τ ) ∂ Cαn(τ )(s )
300 M. Latorre and J.D. Humphrey / Journal of the Mechanics and Physics of Solids 125 (2019) 298–325
Fig. 1. Schematic representation of different configurations involved in soft tissue G&R. The original in vivo homeostatic configuration of the mixture
κ (0 ) = κo is chosen as the reference configuration for the computation of G&R deformations of the mixture via F(τ ), τ ≤ s. Fαn(τ ) (s ) = F(s )F−1 (τ )Gα (τ )
describes the deformation experienced, at time s, by the material element of constituent α deposited at time τ . We assume that the constituents, deposited
with prestretches Gα (τ ), are constrained to deform with the mixture.
where Jnα(τ ) (s ) = det Fα

n (τ )
(s ) = J (s )/J (τ ), with J = det F at time s or τ .
Also let ρ α represent the spatial (apparent) mass density of constituent α (i.e., its mass per unit current volume of
mixture) so that the assumed constant spatial mass density ρ of the mixture is

N
ρ ≡ ρ(s ) = ρ α(s ) . (5)
α =1
The referential mass density ρR (s ) = J (s )ρ (s ) ≡ J (s )ρ of the mixture (i.e., mass per unit reference volume) can similarly
be expressed as

N
ρR(s ) = ρRα(s ), (6)
α =1
with ρRα(s ) = J (s )ρ α(s ). Consistent with Eq. (2), ρRα(s ) reads in constrained mixture models (Latorre and Humphrey 2018b)
s
ρRα(s ) = mαR(τ )qα(s, τ )dτ (7)
−∞
with mαR
(τ ) = J (τ )mα(τ ) the corresponding referential mass density production rate of constituent α . Note from Eqs. (2) and
(7) that one must specify constitutive relations for qα (s, τ ), mα τ , and Wˆ α (Cαn(τ )(s ) ) for the particular soft tissue under
R( )
study.
Following previous studies of arterial G&R (Baek et al. 2007b; Latorre and Humphrey 2018b; Valentín and Humphrey
2009), we let the degradation of structural constituents be described by first-order type kinetics

s
qα(s, τ ) = exp − kα(t )dt (8)
τ
where kα is a rate-type parameter for mass removal that may depend on biomechanical or biochemical factors. For illustra-
tive purposes, let (Latorre and Humphrey 2018b)

kα(t ) = kαo 1 + (σ(t ) )2 , t ∈ [τ , s ] (9)
with kα
o the original homeostatic value and σ the relative difference of a given coordinate invariant measure σ˜ of the
Cauchy stress σ with respect to its original homeostatic value σ˜ o, namely
σ˜ (t ) − σ˜ o
σ(t ) = . (10)
σ˜ o
Moreover, let the mass density production rate mα τ be described by (Latorre and Humphrey 2018b)
R( )
mαR(τ ) = kα(τ )ρRα(τ )ϒ α(τ ) (11)

where kα(τ )ρRα(τ ) represents an evolving nominal mass production rate of constituent α per unit total reference volume
and ϒ α (τ ) is a stimulus function for G&R that can be modulated by biomechanical stimuli (e.g., in healthy arteries; Latorre
and Humphrey 2018a, 2018b; Valentín and Humphrey 2009) or other effects, as, for example, biochemical mediators or
inflammation (Baek et al. 2007b; Latorre and Humphrey 2018c; Miller et al. 2014). Importantly for the present case wherein
we consider mechanobiological stability of previously equilibrated mechanobiological states, mα

R
, as given in Eq. (11), should
ensure a production that balances removal in any evolved homeostatic states “h ” with ϒhα → 1 and hence ρRh α → mα /kα .
Rh h
Of course, ϒoα ≡ 1 and ρRo
α ≡ mα /kα in the original homeostatic state “o”.
Ro o
2.2. Constrained mixture framework: Rate form
Consider now the rate of change of the referential mass density ρRα given in Eq. (7), which yields, by the Leibniz integral
rule,
s
ds ∂ qα (s, τ )
ρ˙ Rα(s ) = mαR(s )qα(s, s ) + mαR(τ ) dτ (12)
ds ∂s
−∞
or, upon consideration of chain and Leibniz rules in Eq. (8), (Latorre and Humphrey 2018b)
ρ˙ Rα(s ) = kα(s )ρRα(s )(ϒ α(s ) − 1 ) (13)
where qα(s, s ) = 1 and we used Eq. (11). Thus, Eq. (13) along with

N
ρ˙ R(s ) = ρ˙ Rα(s ) (14)
α =1
are equivalent to Eqs. (7) and (6), respectively, though in rate form. For later use, since ρR (s ) = J (s )ρ , with ρ constant, note
the rate of change
J(˙ s ) ρ˙ R(s ) N
ρ α(s ) α N
= = kα(s ) R (ϒ (s ) − 1 ) = kα(s )φ α(s )(ϒ α(s ) − 1 ) (15)
J(s ) ρR(s ) α=1 ρR(s ) α =1
with φ α := ρ α /ρ the spatial mass fraction of constituent α , where in this case, ρ α /ρ ≡ ρRα /ρR .
Evolution of the stress given by Eqs. (1) and (2) can also be described by an equivalent rate form. Consider

N
σ˙ (s ) = σ˙ α(s ) − p˙(s )I (16)
α =1
α
where σ˙ yields, from Eq. (2) and the Leibniz rule,
α s s
mα(s )qα(s, s )σˆ (s, s ) ds 1 ∂ qα(s, τ ) α 1 ∂ σˆ α(s, τ )
σ˙ α(s ) = + mα(τ ) σˆ (s, τ )dτ + mα(τ )qα(s, τ ) dτ (17)
ρ ds ρ ∂s ρ ∂s
−∞ −∞
which, with qα(s, s ) = 1, ∂ qα(s, τ )/∂ s = −kα(s )qα(s, τ ), mα(s ) = kα(s )ρ α(s )ϒ α(s ), and a (potentially evolving) deposition
α α
Cauchy stress at the constituent level (cf. Eq. (4)) σˆ (s, s ) = Gα(s )Sˆ α (Gα 2(s ) )Gα(s ) = σˆ dep(s ), reads
s
1 ∂ σˆ α(s, τ )
σ˙ α(s ) = kα(s )ϒ α(s )σ αdep(s ) − kα(s )σ α(s ) + mα(τ )qα(s, τ ) dτ (18)
ρ ∂s
−∞
α
where we define the current deposition stress at the mixture level as σ α
dep
(s ) = φ α(s )σˆ dep(s ).
α
Regarding the third term in the right-hand side of Eq. (18), ∂ σˆ (s, τ )/∂ s yields, from Eq. (4) (with the notation F˙ α
n (τ )( )
s =
α α α
∂ Fn(τ )(s )/∂ s and Jn(τ )(s ) = ∂ Jn(τ )(s )/∂ s)
˙
σ α(s, τ )
∂ 1 1
= α F˙ α (s )Sα Cαn(τ )(s ) Fαn(Tτ )(s ) + α Fα (s ) Sα Cαn(τ )(s ) F˙ αn(Tτ )(s )
∂s Jn(τ )(s ) n(τ ) Jn(τ )(s ) n(τ )
(19)
1 ∂ Sα Cαn(τ )(s ) Jn˙α(τ )(s )
+ α α α
Fn(τ )(s ) Fn(τ )(s ) : − α2 Fαn(τ )(s )
Sα Cαn(τ )(s ) Fαn(Tτ )(s )
Jn(τ )(s ) ∂s Jn(τ )(s )
which involves the rates F˙ α s , J˙α s and ∂ Sˆ α (Cα

n (τ )( ) n (τ )( )
s )/∂ s, where Sˆ α (Cα
n (τ )( )
ˆ α /∂ Cα (s ), and (for further con-
s ) = 2∂ W
n (τ )( ) n (τ )
venience) the symbol represents (A B )i jkl = Aik B jl while the symbol : represents the usual double contraction operation
between second- or higher-order tensors, e.g., A : B = Ai j Bi j . First, from Eq. (3)
∂ Fαn(τ )(s )
F˙ αn(τ )(s ) := = F˙ (s )F−1(τ )Gα(τ ) = l(s )Fαn(τ )(s ) (20)
∂s
where l(s ) = F˙ (s )F−1(s ) is the spatial velocity gradient at the mixture level at current G&R time s. With Jnα(τ )(s ) = J(s )/J(τ ),
we have
∂ Jnα(τ )(s ) J(˙ s ) J(˙ s ) α
Jn˙α(τ )(s ) := = = J (s ) (21)
∂s J(τ ) J(s ) n(τ )
where J(˙ s ) = J(s ) tr l(s). Moreover, the chain rule yields

∂
Sα Cαn(τ )(s ) ∂
Sα Cαn(τ )(s ) ∂ Cαn(τ )(s ) 1
=2 α : : C˙ (s ), (22)
∂s ∂ Cn(τ )(s ) ∂ C (s ) 2
where we identify the referential constitutive (hyperelastic) fourth-order tangent tensor at the constituent level
∂ Sˆ α (Cαn(τ )(s ) ) ∂ 2Wˆ α (Cαn(τ )(s ) )

ˆ α (Cαn(τ )(s ) ) := 2
C =4 α (23)
α
∂ Cn(τ )(s ) ∂ Cn(τ )(s ) ∂ Cαn(τ )(s )
with the symbol representing (A B )i jkl = Ai j Bkl , and the purely kinematic fourth-order tensor (Latorre and Humphrey
2018b)
∂ Cαn(τ )(s )
= Gα(τ )F−T(τ ) Gα(τ )F−T(τ ) . (24)
∂ C (s )
1 −T ˙ −1
Hence, knowing that the rate of deformation tensor d = sym l = 2F CF , we have, from Eq. (22), along with Eqs. (23) and
(24),

∂
Sα Cαn(τ )(s )
α Cα (s ) : Fα T (s ) Fα T (s ) : d(s ).
=C (25)
n (τ ) n (τ ) n (τ )
∂s
cα through the following push-forward operation
Defining a spatial constitutive (hyperelastic) fourth-order tangent tensor
over C α
ˆ at the constituent level
1
cα(s, τ ) = α
Fα (s ) Fαn(τ )(s ) : C
ˆ α (Cαn(τ )(s ) ) : Fαn(Tτ )(s ) Fαn(Tτ )(s ), (26)
Jn(τ )(s ) n(τ )
then Eq. (19), with Eqs. (20), (21), and (25), reads
∂ σˆ α(s, τ ) α α J(˙ s ) α
cα(s, τ ) : d(s ) −
= l(s )σˆ (s, τ ) + σˆ (s, τ )lT(s ) + σˆ (s, τ ) . (27)
∂s J(s )
Finally, substitution of Eq. (27) into Eq. (18) yields the rate equation
σ˙ α (s ) = kα (s )ϒ α (s )σ αdep (s ) − kα (s )σ α (s ) + cα (s ) : d(s ) + l(s )σ α (s ) + σ α (s )lT (s ) − σ α (s )tr l(s ) (28)
α
where we used Eq. (2) to obtain mixture-level stresses σ α (s) from constituent-level stresses σˆ (s, τ ) and, in parallel, we
α α
defined associated mixture-level moduli c in terms of constituent-level moduli c (s, τ ) through the hereditary integral

s
1
cα(s ) := cα(s, τ )dτ .
mα(τ )qα(s, τ ) (29)
ρ
−∞
Eq. (28) reveals multiple contributions to the (instantaneous) change of stress σ α at G&R time s that are intrinsically in-
cluded in Eq. (2), but can only be distinguished in rate form. The first two addends associate, respectively, with an increase
in stress σ α over time due to the deposition at rate kα ϒ α of mass (cf. Eq. (13)) having prestress σ α
dep
and removal at rate
kα of constituent α (cf. Eq. (13)) having stress σ α , hence emphasizing the importance of rates of true production (by de-
position) and removal (by degradation or death). The third addend includes material nonlinearities described constitutively
by the spatial (linearized) stiffness tensor cα . The fourth and fifth addends include well-known geometric nonlinearities
associated with instantaneous changes of the reference configuration for the Cauchy stress (i.e., as given by l, see Eq. (20)).
Finally, the sixth addend highlights a change in stress associated with a change in current total volume (consequently, total
mass) of the soft tissue over G&R timescales, as given constitutively by Eq. (15) in a coupled manner (recall that stimulus
functions ϒ α are typically driven by stress). We assume that such mass addition occurs interstitially due to local cellular
synthesis and secretion.
Remark 1. An important (yet controversial) issue that can arise when developing evolution equations directly in rate form
is selection of an appropriate objective (i.e., frame indifferent) rate in which a spatial quantity (typically, stress) is expressed
and how it relates to kinematic and/or physical quantities and their objective rates (Simó and Pister 1984). In this regard,
α
observe that the expression for σ˙ in Eq. (28) is not objective. Note, however, that we did not posit a constitutive equation
α
for σ˙ ; rather we obtained the (non-objective) material time derivative of the (objective) integral-type expression for σ α
given in Eq. (2). In other words, both equations (Eq. (2), and Eq. (28) including all terms) describe the same evolution for
σ α over G&R time s, hence we can write Eq. (28) as a proper objective rate of the stress tensor σ α . Recalling the Truesdell
rate (cf. Holzapfel 20 0 0, p. 195), here written for each constituent α ,
◦α
σ (s ) = σ˙ α (s ) − l(s )σ α (s ) − σ α (s )lT (s ) + σ α (s )trl(s ) (30)
allows us to write Eq. (28) as
◦
σ α (s ) = kα (s ) ϒ α (s )σ αdep (s ) − σ α (s ) + cα (s ) : d(s ) (31)

which is now an objective equation for the rate of change of mixture level Cauchy stresses σ α whose (convolution-type)
solution is given, equivalently, by Eq. (2). Of course, other objective (e.g., Oldroyd/Lie, Green–Naghdi, or Jaumann) rates
would give different expressions for the same constitutive relation in rate form (i.e., the expanded Eq. (28)). Eqs. (13) and
(31), for example, thus constitute a pair of objective equations in rate form equivalent to Eqs. (7) and (2), respectively, given
in (hereditary) integral form.
Remark 2. Noting that J σ α = FSα FT , with Sα the second Piola–Kirchhoff stress tensor for constituent α at the mixture level
(Latorre and Humphrey 2018b), Eq. (30) can be expressed in terms of the material time derivative of Sα as
◦α 1
σ (s ) = F(s )S˙ α (s )FT (s ) (32)
J (s )
so, from Eq. (31),

1
S˙ α (s ) = kα (s ) ϒ α (s )Sαdep (s ) − Sα (s ) + Cα (s ) : C˙ (s ) (33)
2
where we defined Sα
dep
and Cα through respective pull-back operations over σ α
dep
and cα , namely the Piola transformations
Sαdep (s ) := J (s )F−1 (s )σ αdep (s )F−T (s ) (34)

and
Cα (s ) := J (s )F−1 (s ) F−1 (s ) : cα (s ) : F−T (s ) F−T (s ) . (35)
With Eq. (33) (equivalently, Eq. (31)) written in this way, we define (cf. Eq. (13))
ρ˙ Rα (s ) α
S˙ αg (s ) := kα (s )[ϒ α (s ) − 1]Sα (s ) = S (s ) (36)
ρRα (s )
and

ρ˙ Rα (s ) α
S˙ αr (s ) := −kα (s )ϒ α (s ) Sα (s ) − Sαdep (s ) = − α + k α s
( ) S (s ) − Sαdep (s ) (37)
ρR ( s )
as well as
1
S˙ αe (s ) := Cα (s ) : C˙ (s ) (38)
2
so that Eq. (33) results from the addition of three contributions, growth-type, remodeling / relaxation-type, and elastic-type,
◦
to the stress rate S˙ α (equivalently, σ α in Eq. (31)), namely
S˙ α (s ) = S˙ αg (s ) + S˙ αr (s ) + S˙ αe (s ) (39)
which, having been derived from the general (nonlinear, finite strain) constrained mixture model of Section 2.1, represents
a generalization of the rate-form evolution equation employed in a temporally homogenized constrained mixture model
for G&R (Cyron et al. 2016). Indeed, as done in viscoelasticity, in which strain-like (Latorre and Montáns 2015; Sidoroff
1974) or stress-like (Holzapfel 1996; Simó 1987) internal state variable approaches can be employed to formulate evolution
equations in rate form, one could compute stress evolutions in Eqs. (36)–(38) using appropriate strain- or stress-based
internal variables, bypassing the need to track contributions of individual structurally significant constituents over the mid-
and long-term past history (e.g., via Eqs. (2) and (29)), with consequent savings in computational time and memory. Of
course, different approaches to G&R (integral- or rate-based, with the latter either internal strain- or stress-based) could
lead to different predictions and results and thus must be subjected to experimental validation.
2.3. A constrained mixture model for arterial G&R in rate form
For illustrative purposes, we now specialize the evolution equations for referential mass density and Cauchy stresses
in rate form for a cylindrical artery in maturity that can exhibit active or passive (pseudoelastic) material behaviors. We
consider three main structurally significant constituents, an elastin-dominated amorphous matrix (α = e), oriented smooth
muscle (α = m), and oriented collagen (α = c). We assume that elastin does not turnover during short to moderate periods
of G&R (Wagenseil and Mecham 2009) while smooth muscle and collagen turnover continuously (generally within 4 to 6
months, Humphrey 2002). We further consider active and passive contributions by smooth muscle (Murtada et al. 2017).
Eq. (6), with N = {e, m, c}, reads for this constrained mixture of solid constituents as

e,m,c
ρR ( s ) = ρRα (s ) = ρRe (s ) + ρRm (s ) + ρRc (s ) (40)
α
whereas Eq. (1), under the assumption of an axisymmetric state of stress (with er , eθ , and ez representing unit vectors in
radial, circumferential, and axial directions, and using the compact notation jj → j for second- and fourth-order tensors) such
that

r,θ ,z
σ (s ) = σ j ( s ) e j e j − p ( s )I , (41)
j
reads for each (principal) component
σr (s ) = σre (s ) − p(s ), (42)

e,m,c
σθ ( s ) = σθα (s ) + σθact(s ) − p (s ), (43)
α
and

e,c
σz ( s ) = σzα (s ) − p (s ), (44)
α
where only the elastin-dominated material (which includes effects of glycosaminoglycans, often of much lower mass fraction
than that of elastin) contributes to radial stress, while all three constituents (with separate passive and active terms for
smooth muscle) contribute to circumferential stress and only elastin and collagen contribute to axial stress. Assuming an
axisymmetric deformation, F reads in terms of principal stretches

r,θ ,z
F (s ) = λ j (s ) e j e j (45)
j
where, from Fig. 1, F(0 ) = Fo = I refers to the original homeostatic (loaded) configuration in maturity (cf. Bellini et al. 2014).
Because of the constituent-specific deposition (pre)stretches, σ (0 ) = σ o = 0, in general. The resulting spatial velocity gradi-
−1
ent l = F˙ F is also symmetric, hence
θ ,z
r,
λ˙ j (s )
l (s ) ≡ d (s ) = e e j. (46)
j
λ j (s ) j
Lastly, since the principal directions of Cauchy stress remain constant over G&R time, we can also consider constant deposi-
tion stretch tensors Gα(τ ) = Gα ∀τ (Latorre and Humphrey 2018a; 2018b; Valentín et al. 2013), with associated constituent-
α α α α α
level deposition stresses σˆ = Gα Sˆ α (Gα 2 )Gα = σˆ dep constant as well (i.e., σˆ = σˆ o ≡ σˆ h ).
2.3.1. Elastin
Consistent with prior comments, arterial elastin is assumed to be deposited and cross-linked in the perinatal period and,
due to its long half-life under normal conditions ( > 25 years), not turnover in maturity. Hence, meR (s ) 0 for all s ≥ 0 herein.
Moreover, in the absence of diseases characterized by increased elastolytic activity, qe (s, 0) ࣃ 1 over short-to-modest periods.
Hence, elastin requires a different treatment within this formulation. We thus account for its contribution to the mechanical
state of the artery at the initial time through an equivalent (fictitious, arising largely from prior somatic growth) deposition
stretch tensor Ge(τ = 0 ) = Ge

r,θ ,z
Ge = Gej e j e j (47)
j
such that additional incremental deformations for elastin given by Eq. (3) at time s, with F(τ = 0 ) = I, read
Fe (s ) := Fen(0 ) (s ) = F (s )F−1(0 )Ge(0 ) = F (s )Ge . (48)
The current mass of elastin per unit reference volume of mixture then remains constant over time, namely
ρRe (s ) = ρRe (0 ) = ρ e (0 ) = ρoe , (49)
which in rate form yields
ρ˙ Re (s ) = 0 (50)
consistent with Eq. (13) with neither production nor removal (i.e., (s ) = 0) for s > 0. ke
Under the same assumptions (no production, no removal), it can be shown that components of Cauchy stress for elastin,
Eq. (2), specialize to (Latorre and Humphrey 2018b)
φoe 2
σ je (s ) = φ e (s )λ2j (s )Gej2 Sêj (s ) = λ (s )Gej2 Sêj (s ) j = r, θ , z (51)
J (s ) j
where Sˆ e (Ce (s )) = 2∂ W
ˆ e (Ce (s ))/∂ Ce (s ), with Ce (s ) := Ce (s ). The material time derivative of σ e (s ) in Eq. (51) yields
n (0 ) j
λ˙ j (s ) e λ˙ (s ) J˙ ( s )
σ˙ je (s ) = [2σ je (s ) + cej j (s )] + c jk (s ) k − σ je (s ) (52)
λ j ( s ) k = j λk (s ) J (s )
which, again, is consistent with the general Eq. (28) without production or removal of elastin for s > 0, with cejk ( j, k = r, θ , z)
given in Appendix A. Note that 2σ je + cej j accounts for geometrically and materially nonlinear stiffnesses in direction ej , as
equivalently derived from a “theory of small on large” (Baek et al. 2007a), and that cejk , with j = k, introduces a Poisson-type
coupling between the stress rate σ˙ e and transverse stretch rate λ˙ .
j k
2.3.2. Smooth muscle

We assume that, like collagen, smooth muscle is continuously produced (cell division) and removed (cell apoptosis),
hence associated referential mass densities and Cauchy stresses in rate form are given by the general formulation of
Section 2.2. In contrast to Eq. (50) for elastin, Eq. (13) for smooth muscle reads
ρ˙ Rm (s ) = km (s )ρRm (s )[ϒ m (s ) − 1] . (53)
The deposition stretch tensor for smooth muscle, assumed to be oriented predominantly in the circumferential direction, is
Gm m
θ = Gθ eθ eθ , (54)
hence the rate of change of passive circumferential stress for smooth muscle is, from Eq. (28),

λ˙ (s ) J˙ ( s )
σ˙ θm (s ) = km (s ) ϒ m (s )φ m (s )σˆ θm − σθm (s ) + 2σθm (s ) + cθmθ (s ) θ − σθm(s ) (55)
λθ (s ) J (s )
where σdep
m
|θ (s ) = φ (s )σˆ θ . Note that the transverse-to-axial coupling term cθ z = 0 because smooth muscle is aligned uni-
m m m
directionally (Appendix A).
2.3.3. Collagen
Consistent with prior models (Bellini et al. 2014), we consider a four-family distribution of collagen fibers, one oriented
circumferentially (θ ), one axially (z), and two symmetric diagonally (d). This collection of fiber families accounts for orienta-
tions observed via microscopy as well as difficult to measure cross-links that contribute to the overall anisotropy. The total
referential mass density of collagen is thus
θ
,z,d
ρRc (s ) = ρRic (s ) (56)
i
which, assuming the same removal and production functions and related constants for all four families, satisfies in rate form
(Latorre and Humphrey 2018b)
ρ˙ Rc (s ) = kc (s )ρRc (s )[ϒ c (s ) − 1] . (57)
Circumferential, axial, and symmetric diagonal deposition stretches are
Gcθ = Gcθ eθ eθ , Gcz = Gcz ez ez , (58)
and
Gcd = Gcd ed ed , ed = sin αd eθ + cos αd ez , with αd = ±αo , (59)
hence circumferential and axial components of rates of change of Cauchy stress are, from Eq. (28),

λ˙ (s ) λ˙ z (s ) J(˙ s )
σ˙ θc (s ) = kc (s ) ϒ c (s )φ c (s )σˆ θc − σθc (s ) + 2σθc (s ) + cθc θ (s ) θ + cθc z(s ) − σθc (s ) , (60)
λθ (s ) λz(s ) J (s )
and
λ˙ z (s ) λ˙ (s ) J(˙ s )
σ˙ zc (s ) = kc (s )[ϒ c (s )φ c (s )σˆ zc − σzc (s )] + [2σzc (s ) + czz
c
(s )] + czcθ (s ) θ − σzc (s ) (61)
λz (s ) λθ(s ) J (s )
where (symmetric) coupling stiffness terms cθc z = czcθ persist because of the diagonal fibers (Appendix A) that contribute to
σθc (along with circumferential fibers) and σzc (along with axial fibers).
2.3.4. Active stress

Consider the tensile stress generated by active contraction of smooth muscle cells, which, similar to the passive stress
contribution, can be expressed in the general form
σθact (s ) = φ m (s )σˆ θact (s ) . (62)
Anticipating that we will assess mechanobiological stability of previously mechanobiologically equilibrated states, we as-
sume the following form for σθact (s ) in terms of an active second Piola–Kirchhoff stress (cf., Eq. (51) for elastin, without
considering deposition stretches for the active contribution)
σθact (s ) = φ m (s )σˆ θact (s ) = φ m (s )λ2θ (s )Sˆθact(s ) (63)
which does not incorporate possible time-dependent readjustment of actomyosin filament overlap to optimize the force–
length response (Baek et al. 2007b), as assumed by Valentín et al. (2013). Yet, Eq. (63) allows Sˆθact to depend on a ratio of
vasoconstrictors to vasodilators C, as, for example

σθact (s ) = φ m (s )λ2θ (s )Sˆθact(s ) = φ m (s )λ2θ (s )Sˆ 1 − e−C (s)
2
(64)
with Sˆ a material constant and C ultimately depending on the flow-induced shear stress τ w over the endothelium through
the linearized expression (Valentín and Humphrey 2009)
C (s ) = CB − CS τw (s ), (65)
with CB and CS constants, where τ w , the relative change in τ w with respect to the original homeostatic value τ wo , can
be expressed in terms of associated volumetric blood flow rates (Q, Qo ) and luminal radii (a, ao ) as (Latorre and Humphrey
2018a, 2018b)
τw (s ) − τwo Q (s )a3o
τ w ( s ) = = − 1. (66)
τwo Qo a3 ( s )
Inclusion of the wall shear stress here reminds us that certain components of Cauchy stress can be important mechanobi-
ologically though not important mechanically; note that flow-induced wall shear stress is typically of the order 1.5 Pa while
the pressure-induced in-plane intramural stresses are of the order 150 kPa, yet Pa-order changes in wall shear stress can
dramatically affect matrix turnover and the geometry in which such turnover occurs. Nevertheless, the material time deriva-
tive of σθact in (63) yields, with φ m = ρ m /ρ ≡ ρRm /ρR ,

ρ˙ Rm (s ) ρ˙ R (s ) act λ˙ (s ) dSâct(C (s ) )
σ˙ θact (s ) = − σθ (s ) + 2σθact (s ) θ + φ m (s )λ2θ (s ) θ C˙ (s ) (67)
ρR ( s ) ρR ( s )
m
λθ (s ) dC(s )
where C˙ reads, from Eqs. (65) and (66),

τ˙ w (s ) τw (s ) a˙ (s ) Q˙ (s )
C˙ (s ) = −CS = CS 3 − . (68)
τwo τwo a (s ) Q (s )
Let an active stiffness-like term cθact
θ > 0 be
τw(s ) dSˆθact(C(s ) )
cθactθ (s ) := φ m (s )λ2θ (s )3CS (69)
τwo dC (s )
and the circumferential stretch for active stresses be approximated by

a (s )
λθ (s ) ≈ (70)
ao
so a˙ /a ≈ λ˙ θ /λθ , such that Eq. (67) adopts a similar expression to its passive counterpart, Eq. (55),

σ˙ θact (s ) = km (s ) ϒ m (s )φ m (s )σˆ θact (s ) − σθact(s )

λ˙ (s ) J˙ ( s ) cact (s ) Q˙ (s )
+ 2σθact (s ) + cθactθ (s ) θ − σθact (s ) − θθ . (71)
λθ (s ) J(s ) 3 Q (s )
Note that an increase (or decrease) in blood flow would potentially lead to an instantaneous vessel dilatation (or constric-
tion) via the relaxation (or further contraction) of its smooth muscle, as desired during acute responses to altered flow
(Humphrey 2002).
2.3.5. Mixture-level constitutive relations

Noticing that we include constituents that either turnover or not and that can have different passive and active contribu-
tions as well as different spatial arrangements and orientations, the constitutive relations for total referential mass density
and Cauchy stresses in rate form specialize for our prototypical artery as

m,c
ρ˙ R (s ) = ρ˙ Rα (s ) = J˙ (s )ρ , (72)
α
and

e,m,c,act
e,c
σ˙ θ (s ) = σ˙ θα (s ) − p˙ (s ), σ˙ z (s ) = σ˙ zα (s ) − p˙ (s ), σ˙ r (s ) = σ˙ re(s ) − p˙ (s ) (73)
α α
where, for notational compaction, we include the superscript act for active circumferential stresses within the summation
over the “constituent” index α . Let the “extra” part of stress (Humphrey 2002) given by different constituents in the principal
directions be denoted with superscript x as

e,m,c,act
e,c
σθx (s ) = σθα (s ), σzx(s ) = σzα (s ), and σrx (s ) = σre (s ) (74)
α α
with respective stiffnesses as

e,m,c,act
e,c
cθx θ (s ) = cθαθ (s ), x
czz(s ) = α ( s ),
czz x
and crr (s ) = crre (s ) (75)
α α
as well as (with cxjk = ckx j )

e,c
cθx z (s ) = cθαz (s ), x
czr(s ) = czr
e
( s ), and cθx r (s ) = cθe r (s ) . (76)
α
Hence, with Eqs. (53) and (57) for referential mass densities, and Eqs. (52), (55), (60), (61), and (71) for stress contributions,
we can write constrained mixture equations in rate form for an idealized cylindrical artery as

m,c
ρ˙ R = kα ρRα (ϒ α − 1 ) (77)
α
θ ,z
cact Q˙
m,c,act
ρα λ˙
r,
λ˙ j ρ˙ R
σ˙ θ + θ θ = kα ϒ α R σˆ θα − σθα + 2σθx θ + cθx j − σθx − p˙ (78)
3 Q α
ρR λθ j
λj ρR
θ ,z
ρc λ˙
r,
λ˙ ρ˙
σ˙ z = kc ϒ c R σˆ zc − σzc + 2σzx z + czx j j − σzx R − p˙ (79)
ρR λz j
λ j ρR
θ ,z
λ˙ r r, λ˙ j ρ˙ R
σ˙ r = 2σrx + crx j − σrx − p˙ (80)
λr j
λ j ρR
where the rates in the right-hand sides are yet to be related to the variables of interest for each particular case. Note that
these, and all relations in Section 2.3, hold at any point within a cylindrical arterial wall whether thin- or thick-walled.
2.4. Equivalent thin-walled artery: Stability analysis
We seek to analyze the mechanobiological stability of evolving loaded states of an artery, which we assume to have
achieved a state of mechanobiological equilibrium when subjected to an inner pressure P = Ph , flow rate Q = Qh , and ax-
ial stretch λz = λzh . Because these external loads are eventually sustained over time during the dynamic stability analy-
ses performed, the resulting solutions represent self-excited deformations caused by combined growth, remodeling, and
elastic responses. That is, we are not analyzing forced dynamic responses, either over short (cardiac cycle) or long (G&R)
timescales. Moreover, because of residual stresses in maturity (which arise from materially non-uniform deposition stretches
and somatic growth), the mean values of Cauchy stress represent well the transmural distribution of stress. Hence, we
use mean values similar to those for a thin-walled pressure vessel. Because of the aforementioned order of magnitude
difference in mean in-plane (θ and z) and out-of-plane (r) stresses, we also assume a quasi-plane-stress state for which
|σ r |/σ θ ∼ |σ r |/σ z 1, hence the Lagrange multiplier in Eqs. (78)–(80) is obtained directly from Eq. (42) as (hereafter, we
omit any dependences on times s or τ for notational convenience)
σr = σre − p = 0 ⇒ p = σre . (81)
Noticing from Eq. (51) that σ je ∝ Gej 2 and that in-plane deposition stretches for elastin are typically Ge
θ ∼ Gez ∼ 2, with Ger =
1/(Geθ Gez ) ∼ 1/4, we have
3
|σre | |σre | 1
∼ ∼ 1 (82)
σθe σze Geθ Gez
and because smooth muscle and/or collagen contribute to circumferential and axial stresses through Eqs. (74)1 and (74)2 ,
we find
| p| = |σre | σθe , σze < σθx , σzx (83)
whereby p will be negligible in this mechanobiological stability analysis. If this assumption did not hold at a given home-
ostatic (loaded) state, a similar procedure could be followed while incorporating p˙ = σ˙ re , from Eq. (80), in Eqs. (78) and
(79). Finally, for analytical convenience, we consider the same (original) rates and gains for turnover of smooth muscle and
collagen, thus
km c
o = ko ≡ ko , and ϒ m = ϒ c ≡ ϒ . (84)
2.4.1. Nonlinear, non-autonomous system

The previous assumptions, along with the external loads remaining constant over time during the stability (self-
excitation) analysis, that is,
P = Ph , Q = Qh , λz = λzh ⇒ P˙ = 0, Q˙ = 0, λ˙ z = 0 (85)
reduce Eqs. (77)–(80) to the following nonlinear system of first-order differential equations
ρ˙ R = k(ϒ − 1 )(ρR − ρoe ) (86)

m,c,act
ρRα α λ˙ λ˙ ρ˙
σ˙ θ = k ϒ σˆ θ − σθ + (2σθ + cθ θ ) θ + cθ r r − σθ R
α (87)
ρ R λ θ λ r ρR
α c
ρ λ˙ λ
˙ ρ˙
σ˙ z = k ϒ R σˆ zc − σzc + czθ θ + czr r − σz R (88)
ρR λθ λr ρR
which suggests that ρ R , σ θ , and σ z could represent an appropriate set of time-dependent variables for an asymptotic sta-
bility analysis of arterial G&R if the three right-hand sides could be expressed as functions of these variables and, per-
haps, G&R time s, either implicitly or explicitly. In other words, we seek to identify a non-autonomous system of the form
(Rouche et al. 1977)
y˙ (s ) = f(y(s ), s ), s ≥ 0, y ( 0 ) = yh + δ yh (89)
with y = [ρR , σθ , σz ] the dependent variable, s the independent variable, f(y, s) a nonlinear vector-valued function, and δ yh
T
an initial (typically modest) perturbation relative to the evolved homeostatic state yh = [ρRh , σθ h , σzh ]T .
Consider first the global equilibrium equation for mean circumferential stress σθ = Pa/h, with transmural pressure P,
luminal radius a, and wall thickness h, which can be expressed in terms of γh = Ph /Po, σθ o = Poao/ho, λθ = a/ao, and ρR /ρ =
J = λr λθ λzh , with λr = h/ho, as
Ph aho λ λ2 λzh λ2
σθ = σθ o = γh σθ o θ = γh σθ o θ = γh λzh ρσθ o θ (90)
Poaoh λr J ρR
Hence, internal (constitutive) and external (mechanical equilibrium) expressions for σ˙ θ yield, from Eqs. (87) and (90)

m,c,act
ρα λ˙ λ˙ ρ˙ λ˙ ρ˙
k ϒ R σˆ θα − σθα + (2σθ + cθ θ ) θ + cθ r r − σθ R = σθ 2 θ − R (91)
α
ρR λθ λr ρR λθ ρR
whereupon the associated (total) Truesdell rate vanishes, cf. Eq. (31),

◦
m,c,act
ρRα α λ˙ λ˙
σθ = k ϒ σˆ θ − σθ + cθ θ θ + cθ r r = 0
α (92)
α
ρR λθ λr
Furthermore, J˙ = J tr l, with Eqs. (15)1 and (46), and λ˙ z = 0, yields
λ˙ θ λ˙ r ρ˙ R
+ = (93)
λθ λr ρR
which, along with Eq. (92), enable expressions for λ˙ θ /λθ and λ˙ r /λr to be substituted into Eqs. (87) and (88). Letting the
stimulus function ϒ in Eqs. (86)–(88) be driven by relative intramural σ (same as in Eq. (9)) and wall shear τ w stresses
(same as in Eq. (65)) through a linearized relation (Latorre and Humphrey 2018b; Valentín and Humphrey 2009), we have
ϒ = 1 + Kσ σ − Kτ τw (94)
with Kσ ≥ 0 and Kτ ≥ 0 gain-type G&R parameters, and σ˜ in Eq. (10) the first principal invariant of σ , namely σI = σr + σθ +
σz σθ + σz , so
σI − σIo σ + σz
σ = θ −1 (95)
σIo σθ o + σzo
which, importantly, accounts for the biaxial wall mechanics, noting that axial mechanics plays fundamental roles in arterial
mechanics (Humphrey et al. 2009), including arterial G&R (Gleason et al. 2007). In addition, we can rewrite τ w /τ wo in
Eq. (66), with εh = Qh /Qo and Eq. (90), as
τw Q a3 ε ρσθ o 3/2
= h 3o = h3 = εh γh λzh (96)
τwo Qoa λθ ρR σ θ
m,c,act
so we obtain the desired dependences for ϒ = ϒ (y ) ≡ ϒ (ρR , σθ , σz ). Now, let the terms α σθα and σzc in Eqs. (87) and
(88) be expressed as

m,c,act
σθα = σθ − σθe , and σzc = σz − σze . (97)
α
If, as in previous work (Latorre and Humphrey 2018a; 2018b; 2018c), the hyperelastic response of elastin is modeled via a
neoHookean function with shear modulus ce , for which Sˆθe = Sˆze = ce are constant, we have from Eqs. (51) and (90)
λ2θ λ 2
σθeo
σθe = φoe Geθ2 Sˆθe = σθeo θ = σ (98)
J J γh λzh σθ o θ
and
λ2zh λ2zh ρ
σze = φoe Gez2 Sˆze = σzo
e
= λ2zh σzo
e
(99)
J J ρR
whereby terms in Eq. (97) can be expressed as a function of the variables in y = [ρR , σθ , σz ]T as

m,c,act
σθeo ρσ e
σθα = 1 − σθ , and σzc = 1 − λ2zh zo σz . (100)
α
γh λzh σθ o ρR σ z
Likewise, because cejk = 0 for this particular case, cθ r = 0 = czr in Eqs. (87) and (88). Consider, finally, the terms in
Eqs. (87) and (88)

m,c,act
ρRα α ρc
σˆ θ , and R σˆ zc (101)
α
ρR ρR
which, after noticing that if (Latorre and Humphrey 2018b)
ρRm ρRc ρRm + ρRc ρR − ρoe
o = ko , ϒ = ϒ
km c m c
⇒ = = = (102)
ρom ρoc ρom + ρoc ρ − ρoe
then
ρRc c ρ ρRc ρoc c ρ ρR − ρoe c
σˆ = σˆ = σ (103)
ρR z ρR ρoc ρ z ρR ρ − ρoe zo
c = φα σ
with the (original) total axial stress for collagen σzo c
o ˆ z constant. Moreover

m,c,act
ρRα α m,c,act
ρ ρα ρα
α = ρ ρR − ρo

e m,c,act
σˆ θ = R
α ρ
o
σ
ˆ θ φoα σˆ θα (104)
α
ρR α
ρ ρ
R o ρ R ρ − ρ e
o α
with

m,c,act
ρR σθ âct
φoα σˆ θα = φom σˆ θm + φoc σˆ θc + φom σˆ θact = σθmo + σθco + φom S (σ , ρ ) (105)
α
γh λzh ρσθ o θ θ R
where σθmo and σθco are constant. Note, from Eqs. (63)–(65), (90) and (96), the dependences for σˆ θact
σˆ θact = λ2θ (σθ , ρR )Sˆθact (σθ , ρR ) . (106)

Therefore, after some lengthy (but otherwise straightforward) algebra, Eqs. (86)–(88) reduce to the following nonlinear
system of differential equations of the form in Eq. (89), which, for prescribed constant loads Ph , Qh , λzh , and initial pertur-
bation δ yh , describes the evolution of ρ R , σ θ , and σ z in terms of themselves and the G&R-time-dependent stiffnesses cθ θ
and czθ (see Eqs. (75), (76), and Appendix A), namely
ρ˙ R ρR − ρoe
= (ϒ − 1 ) (107)
kρR ρR
σ˙ θ ρR − ρoe 2σ
=− (ϒ − 1 ) + θ (108)
kσθ ρR cθ θ
σ˙ z ρR − ρoe 2σ c
=− ( ϒ − 1 ) + θ zθ − χ (109)
kσz ρR c θ θ 2 σz
where
σ +σ τ
θ z w
ϒ = 1 + Kσ − 1 − Kτ −1 (110)
σθ o + σzo τwo
and we defined

σθeo 1 ρR − ρoe ρ σθmo + σθco 1 φom Sˆθact
:= 1 − −ϒ + , (111)
γh λzh σθ o ρ − ρoe ρR σθ γh λzh σθ o
and
ρ σzoe ρR − ρoe ρ σzoc
χ := 1 − λ2zh −ϒ (112)
ρR σ z ρ − ρoe ρR σz
with τ w /τ wo (in ϒ ) and Sˆθact (in ) functions of ρ R and σ θ from Eqs. (96) and (106). Although one would need to update the
hereditary-integral-based elastic moduli cθ θ and czθ (cf. Eq. (29)) between incremental steps to solve numerically Eqs. (107)–
(109) for ρ R , σ θ , and σ z , we will just need their homeostatic values cθ θ h and czθ h to analyze the stability of the dynamical
system of Eqs. (107)–(109) near a homeostatic state.
2.4.2. Mechanobiological equilibrium

As mentioned above, we seek a mechanobiological stability analysis of an equilibrated mechanobiological state, previously
computed for prescribed Ph , Qh , λzh . Mathematically, the equilibrated state yh must satisfy ρ˙ R = 0, σ˙ θ = 0, and σ˙ z = 0 in
Eqs. (107)–(109), that is, from Eq. (89)
0 = f(y(s ), s ) ∀s ≥ 0 ⇒ y ( s ) = yh (113)
hence representing a so-called equilibrium or critical point (Rouche et al. 1977) of this system of first-order differential
equations.
Considering ρ˙ R |h = 0, σ˙ θ |h = 0, and σ˙ z |h = 0 in Eqs. (107)–(109), requires
ϒh = 1 , h = 0 , and χh = 0 (114)
at mechanobiological equilibrium. It can be shown, numerically, that the steady-state solution yh = [ρRh , σθ h , σzh ] obtained T
from Eq. (114), via our rate-based constrained mixture approach, represents the same mechanobiologically equilibrated solu-
tion obtained from an integral-based constrained mixture approach for the same prototypical vessel (Latorre and Humphrey
2018b). In particular, it is easy to verify, analytically, that the original homeostatic rule-of-mixture solution (cf. Section 3.2 in
Latorre and Humphrey (2018b), with volume and mass fractions being equivalent and p negligible with respect to in-plane
stresses)

e,m,c
ρR = ρoα ≡ ρ , (115)
α

e,m,c,act
e,m,c,act
σθ = φoα σˆ θαo = σθαo ≡ σθ o , (116)
α α
and

e,c
e,c
σz = φoα σˆ zoα = σzoα ≡ σzo (117)
α α
obtained for P = Po, Q = Qo, and λz = λzo = 1, satisfies ϒo − 1 = o = χo = 0 in Eqs. (110)–(112), and, thus, represents an
associated critical point of the system of first-order Eqs. (107)–(109).
2.4.3. Mechanobiological stability

Importantly, distinction between time-dependent and time-independent solutions, Eqs. (89) and (113), respectively, al-
lows us to analyze two types of (un)stable responses related to different mechanobiological sources. Briefly, Eq. (89) de-
scribes the evolution of y(s) near a previously mechanobiologically equilibrated solution yh following an initial perturbation
δ yh at s = 0 for prescribed model parameters and external loads for s ≥ 0, and, hence, requires that the critical point yh ex-
ists and is bounded, in general. Subsequently, a dynamic stability analysis determines if the time-dependent solution y(s > 0)
approaches yh , remains close to yh , or diverges. On other hand, Eq. (113) yields a time-independent mechanobiologically equi-
librated solution yh for prescribed model parameters and external loads, which might be statically bounded or unbounded,
or even give rise to bifurcations. We thus refer to mechanobiological dynamic stability as the ability of the time-dependent
solution y(s) to remain close to a finite (bounded) equilibrated state yh , with rate-dependent terms in Eq. (89) playing a
central role. We alternatively refer to mechanobiological static stability as the ability of time-independent solutions yh to
remain finite (bounded), where only rate-independent terms in Eq. (89) are relevant.
2.4.3.1. Mechanobiological (static) stability of yh with respect to sustained perturbations. Consider our idealized artery with
initial geometry and mass fractions xo = {ao, ho, lo, φo} in an original homeostatic (loaded) state o. An associated original
equilibrated solution yo = [ρRo, σθ o, σzo]T (with ρRo = ρ ) is obtained from Eq. (113) for prescribed original external loads
ξ o = {Po, Qo, λzo} and original model parameters ς o , namely, Eqs. (115)–(117). Consider, too, an evolved equilibrated solution
yh = [ρRh , σθ h , σzh ]T , obtained from Eq. (113) for evolved loads ξ h = {Ph , Qh , λzh } and evolved parameters ς h , which can be
computed numerically from the three conditions in Eq. (114) (cf., equivalently, Latorre and Humphrey 2018b). The steady-
state solution yh is called mechanobiologically stable with respect to sustained changes in external loads ξ h if, for every
(physiological) > 0, there is a (physiological) δ > 0, such that

ξ h − ξ o < ⇒ yh − yo < δ. (118)
Equivalently, yh is called mechanobiologically stable with respect to sustained changes in parameters ς h if, for every (phys-
iological) > 0, there is a (physiological) δ > 0, such that
ς h − ς o < ⇒ yh − yo < δ. (119)
Otherwise, the steady-state solution yh is mechanobiologically unstable with respect to physiologically admissible changes
in ξ h or ς h . Note that neither G&R time s nor rate-dependent terms influence this type of statically (un)bounded solution.
Eq. (113) with y(s ) = yh , absent G&R time s, and explicit consideration of the evolution of a single parameter ζ h ∈ {ξ h ,
ς h } can be rewritten as a one-parameter family of equations f(yh , ζh ) = 0. By the implicit function theorem, we have
f(yh , ζh ) = 0 ⇒ yh = yh (ζh ), (120)
hence static stability of yh with respect to changes in ζ h can be assessed through the evolution of the (implicit) parameter-
dependent solution yh = yh (ζh ). Differentiation of f(yh (ζh ), ζh ) = 0 with respect to ζ h yields
df(yh (ζh ), ζh ) ∂ f(yh , ζh ) dyh (ζh ) ∂ f(yh , ζh )

= · + =0 (121)
dζh ∂ yh dζh ∂ζh
whereupon
−1
dyh (ζh ) ∂ f(yh , ζh ) ∂ f(yh , ζh )
=− · . (122)
dζh ∂ yh ∂ζh
A first-order Taylor expansion yields yh (ζh + ζh ) yh (ζh ) + (dyh (ζh )/dζh )ζh . Assuming that yh increases monotoni-
cally with respect to monotonic changes in ζ h , mechanobiological static stability of yh , in the sense of Eqs. (118) or (119),
requires, from Eq. (122), that the Jacobian matrix ∂ f(yh , ζ h )/∂ yh is invertible. In other words, if there exists a physiological
value ζ h (i.e., |ζh − ζo| < ), such that

∂ f(yh , ζh )
det → 0 ⇒ yh (ζh ) − yo → ∞ (123)
∂ yh
then the mechanobiologically equilibrated solution yh is statically unstable (i.e., unbounded). For our idealized artery,
⎡ ⎤
∂ ρ˙ R ∂ ρ˙ R ∂ ρ˙ R
⎢ ∂ρR
⎢ ∂σθ ∂σz ⎥
⎥
∂ f(yh , ζh ) ∂ f(y, ζ ) ⎢ ∂ σ˙ θ ∂ σ˙ θ ∂ σ˙ θ ⎥
∂ y h ⎢ ⎥
= = (124)
∂ yh ⎢ ∂ρR ∂σθ ∂σz ⎥
⎣ ∂ σ˙ ∂ σ˙ z ∂ σ˙ z ⎦
z
∂ρR ∂σθ ∂σz h
which we compute numerically using Eqs. (107)–(109) in examples below. Importantly, other types of G&R instabilities (not
necessarily unbounded) could arise depending on the specific evolution of the nonlinear function yh (ζ h ), including limit-
point instabilities and/or bifurcations (cf. Erlich et al. 2018). Hence, each case, defined by specific constitutive relations,
material constants, geometry, and boundary conditions should be evaluated separately (Haslach and Humphrey 2004).
2.4.3.2. Mechanobiological (dynamic) stability of y(s) with respect to transient perturbations near yh . As explained above, a
mechanobiologically equilibrated solution yh represents a critical point of Eq. (89) for a given original geometry and mass
fractions xo , prescribed evolved loads ξ h , and prescribed evolved model parameters ς h . Assume now that for a given h > 0,
there exists a δ h > 0 such that Eqs. (118) and (119) are satisfied, or, in other words, that yh exists and remains physiological.
Eq. (89), with xo given, and ξ h and ς h fixed, describes a time-dependent solution y(s ≥ 0) following an initial perturba-
tion δ yh at s = 0. The time-dependent solution y(s) is called mechanobiologically stable at s = 0 with respect to arbitrary
perturbations δ yh if, for every > 0, there is a δ > 0 such that if
δ y h < ⇒ y ( s ) − y h < δ ∀ s ≥ 0 . (125)
Otherwise, y(s) is mechanobiologically unstable at s = 0 with respect to transient perturbations δ yh . Note that both the G&R
time s and rate-dependent terms play crucial roles in dynamically (un)bounded solutions.
Consider an evolved equilibrium solution yh = [ρRh , σθ h , σzh ]T of Eqs. (107)-(109) for prescribed ξ h and ς h , with y˙ = 0.
We linearize about yh (0), as given in Eq. (89), as

∂ f(y, s ) ∂ f(y, s )
y˙ = f(y, s ) f(yh , 0 ) + · ( y − y ) + (s − 0 ) (126)
∂ y h h
∂ s h
where we neglect higher-order terms. Considering that cθ θ and czθ are additional variables (to ρ R , σ θ , and σ z ) that depend
on s,
⎡ ⎤ ⎡ ⎤
∂ ρ˙ R 0
⎢ ∂s ⎥ ⎢ 2σ dc ⎥
∂ f(y, s ) ⎢ ⎥ ⎢ − 2 θ θθ ⎥
= ⎢ ∂ σ˙ θ ⎥ = h ⎢ cθ θ ds ⎥ =0 (127)

∂s h ⎣ ∂s ⎦ ⎢ ⎥ ⎢ ⎥
⎣ 2σ c2 d (c /c ) ⎦
∂ σ˙ z − 2 θ zθ θ θ zθ
∂s h c θ θ 2 σz ds
h
which, importantly, vanish because of the equilibrium value h = 0 in Eq. (114). Thus, since yh is an equilibrium point (i.e.,
f(yh , 0 ) = 0) and ∂ f(y, s )/∂ s|h = 0, linearization of Eq. (89) at yh (i.e., Eq. (126) is represented by a linear autonomous system
of differential equations in terms of an incremental (time-dependent) solution δ y(s ) = y(s ) − yh

∂ f(y, s )
δ y˙ (s ) = · δ y (s ) , s≥0, δ y ( 0 ) = δ yh , (128)
∂y h
which is valid in a neighborhood of the evolved homeostatic solution y = yh .

It is well-known (Hairer et al. 1993) that the eigenvalues of the matrix of constant coefficients ∂ f(y, s)/∂ y|h determine the
stability of the associated linear(ized), autonomous system given in Eq. (128). The question now is whether these eigenvalues
determine, too, the stability of the nonlinear, non-autonomous system given in Eq. (89) near the equilibrium state yh . This
theory, initiated in the late 1800s, was “brought to perfection” (Hairer et al. 1993) by Lyapunov (1882), so we employ his
main contributions in this regard.
2.4.3.3. Original homeostatic state. For illustration, we analyze analytically the asymptotic stability of y(s) near the original
equilibrium state yo . After considering the dependences of the right-hand sides of Eqs. (107)–(109) on ρ R , σ θ , and σ z ,
performing all the required partial derivatives in Eq. (124), and rearranging terms conveniently (see Appendix B), we obtain
the following (generally complex) eigenvalues for the linearized problem of Eq. (128) particularized at the original state yo
l1o = −ko (129)

ko

l2o = To + To2 − 4Do (130)
2
ko

l3o = To − To2 − 4Do (131)
2
where To := (tr (∂ f/∂ y|o ) − l1o )/ko = tr (∂ f/∂ y|o )/ko + 1 reads

2σθneo Cθ zo σˆ ne 2σθneo C act 2σθneo
To = − + φone − φoe zo K̄σ − K̄τ − θ θ o + , (132)
c θ θ o 2 σθ o σθ o cθ θ o cθ θ o cθ θ o
and Do := (det(∂ f/∂ y|o )/l1o )/k2o = − det(∂ f/∂ y|o )/k3o reads

2σθneo σˆ ne Cθact
θo eσ
ne
ˆ zo 2σθneo
Do = φ e
1 + zo + φo − φo
ne
K̄σ + K̄τ , (133)
cθ θ o o
σθ o cθ θ o σθ o cθ θ o
with K̄σ = Kσ σθ o/σIo, K̄τ = 3Kτ /2, φone = 1 − φoe , σθneo = σθ o − σθeo, σˆ zo
ne = σ − σ
zo
e, C
ˆ zo θ zo = 2σθ o + cθ zo, and Cθ θ o = 2σθ o + cθ θ o.
act act act
Clearly, many parameters influence the system, including mass fractions, active stress and stiffness (which can depend on
vasoactive parameters), passive circumferential and axial stresses, and circumferential and axial-to-circumferential total stiff-
nesses (that depend, additionally, on elastic constants, deposition stretches and diagonal collagen orientation), and, of course,
rate and gain parameters for removal and production.
Regarding the type and sign of the eigenvalues l1o to l3o in Eqs. (129)–(131), we firstly observe that
ko ≷ 0 ⇒ l1o ≶ 0 . (134)
Hence, for l1o < 0 (i.e., with ko > 0, which is physical), one can assure, based on a linearized analysis (Rouche et al. 1977),
that the following (partial) results regarding the critical point yo = [ρ , σθ o, σzo]T of the nonlinear system of Eqs. (107)–(109)
hold in a neighborhood of yo (cf. the Trace-Determinant plane, Figure 9.1.9 in Boyce and DiPrima 2012) such that
To < 0 and Do > 0 ⇒ y(s ) is asymptotically stable (135)
or
To > 0 or Do < 0 ⇒ y(s ) is unstable (136)
where neglect of higher-order terms in Eq. (126) intentionally disregarded particular cases that would generally require
their consideration (e.g., complex eigenvalues with vanishing real parts, giving rise to “the [nonlinear] center problem”,
cf. Hairer et al. 1993). Importantly, if, because of the current lack of empirical evidence of oscillatory behaviors in vivo
over G&R timescales, one assumes that physiological G&R responses are dynamically stable and progress over time without
oscillations, then To < 0 and Do > 0 in Eq. (135) must additionally satisfy

To < −2 Do < 0 ⇒ y(s ) is asymptotically stable without oscillations (137)
which, based on Eqs. (132) and (133), imposes conditions on mechanobiological parameters to ensure physiologically reason-
able dynamical adaptations. Indeed, depending on different values of the trace To and determinant Do (hence, eigenvalues
l2o and l3o ), we found in numerical examples below that so-called asymptotically stable (spiral or nodal) sinks, unstable
(spiral or nodal) sources, or neutrally stable centers (Boyce and DiPrima 2012) are mathematically admissible.
2.5. Illustrative results for the murine aorta
Recall that the mechanical response of elastin is modelled using a neoHookean relation
ce e
ˆ e (Ce (s )) =
W (C (s ) : I − 3 ), (138)
2
with ce a shear modulus, which we used to express σθe and σze in Eqs. (98) and (99) in terms of the primary variables σ θ
and ρ R . Conversely, smooth muscle and collagen are modelled using Fung-type relations
c1α

α α2

ˆ α (λαn τ (s )) = ec2 (λn(τ ) (s )−1) − 1 ,
2
W α = m, c, (139)
( ) 4c2α
where c1α (dimensions of stress) and c2α (dimensionless) are material parameters, and λαn (τ )
(s ) is the fiber stretch (relative to
its evolving natural configuration due to continued matrix production). Collagen fiber families oriented in circumferential,
axial, and symmetric diagonal directions have respective fractions β θ , β z , and βd = 1 − βθ − βz . Representative values of
parameters for a mouse descending thoracic aorta are listed in Table 1, with smooth muscle and collagen sharing rate and
gain constants for convenience, recall Eqs. (84) and (94). Additional values needed to quantify the active response of smooth
muscle through Eqs. (64) and (65) are given in specific examples below. The inner pressure at the original homeostatic
state, for vanishing active contribution, is Po = 13.2 kPa, with mean values of circumferential σθ o = 213 kPa and axial σzo =
238 kPa stresses in the original homeostatic state, consistent with experimental findings (Bellini et al. 2014).
Fig. 2. Mechanobiologically stable (static) equilibrium responses illustrated for a “normal” murine aorta. Panels (a,b,d,e) show equilibrium values for
(bounded) inner radius ah and thickness hh , as well as (bounded) circumferential σ θ h and axial σ zh Cauchy stress, as functions of the stimulation-driver
pressure ratio γh = Ph /Po . Panels (c) and (f) show the associated evolution of the homeostatic state in phase-type planes: thickening with slight dilation and
slight reduction in biaxial stress. Note, too, that the thickening is not fully mechano-adaptive, consistent with experimental observations for the murine
thoracic aorta (Bersi et al. 2016).
3. Illustrative results
Here, we consider three cases of (patho)physiological importance – an acute but sustained increase in blood pressure, a
pathological loss of elastin, and the role of active smooth muscle contraction – as well as two studies of key constitutive
parameters.
3.1. Mechanobiological stability of equilibrium solutions
3.1.1. Acute increase in pressure

Consider the equilibrium inner pressure ratio γh = Ph /Po as the driving parameter ζ h for a one-parameter family of non-
linear equations of the type (120)1 , explicitly, Eqs. (107)–(109) with y˙ = [ρ˙ R , σ˙ θ , σ˙ z ]T = 0 ∀s ≥ 0, or, alternatively, Eqs. (114).
Let the flow rate and axial stretch preserve their original homeostatic values εh = Qh /Qo = 1 and λzh = 1. The solution to
these time-independent equations, with baseline parameters in Table 1, represents a mechanobiologically equilibrated state
for each γ h , that is, yh (γ h ) in Eq. (120)2 . Corresponding inner radius ah , thickness hh and axial force fh are then obtained
from ρRh /ρ = Jh plus the mechanical equilibrium equations σθ h = Ph ah /hh and σzh = fh /(2π ah hh ), with Jh /λzh = λrh λθ h =
hh ah / ( ho ao ).
Fig. 2 shows equilibrium values for (a) ah , (b) hh , (d) σ θ h , and (e) σ zh as functions of the stimulation-driver Ph /Po . In
addition, panels (c) and (f) show the evolution of these variables in respective phase-type planes, with the driving pa-
rameter γ h removed. The pressure was increased from its original homeostatic value γo = 1 (i.e., Po ≈ 99 mmHg) up to
an evolved ratio γh = 1.8 (i.e., Ph ≈ 178 mmHg), hence = 0.8Po in Eq. (118), which is a proper range of biological inter-
est. Observe that all the equilibrium variables remain statically bounded for increasing pressure within this range (follow-
ing, e.g., Humphrey 2002, qualitatively), which, according to Eq. (118), represents a mechanobiologically stable situation.
The problem remains well-posed for this range of pressures, with the γ h -dependent Jacobian determinant, cf. Eq. (123),
Do = − det(∂ f/∂ y|o )/k3o = 0.15 for γo = 1 and Dh = − det(∂ f/∂ y|h )/k3h = 0.10 for γh = 1.8. The partial derivatives ∂ f/∂ y|h at
equilibrium points yh (γ h ), see Eq. (124), were computed numerically via forward (first-order) finite differences. In partic-
ular, the Jacobian determinant computed numerically at the original homeostatic state was consistent with the analytical
expression in Eq. (133). Results show wall thickening with slight dilatation as is common in hypertension (Humphrey 2002).
3.1.2. Elastin degradation

Consider G&R driven, quasi-statically, by elastin degradation prescribed by ζh ≡ che = (1 − ϕh )coe , where coe is the original
shear modulus for elastin, che is the evolved modulus used to compute the associated equilibrated states, and ϕh ∈ [0, 1]
quantifies elastin degradation. In addition to the flow rate and axial stretch, the inner pressure also remains constant, that
is, γh = εh = λzh = 1. Note that the modified elastin properties, Eqs. (98) and (99) derived for a constant value ce , needed to
be updated by the corresponding factor, che /coe = 1 − ϕh . If Kτ = Kσ /2 (Table 1), we find a bounded growth of the vessel for
all ϕh ∈ [0, 1]; in contrast, if Kτ = Kσ /5, we find an unbounded growth at some ϕ h < 1. Specifically, Fig. 3 shows equilibrium
Fig. 3. Mechanobiologically unstable (static) equilibrium responses with respect to prior elastin degradation while preserving inner pressure Ph = Po . Panels
(a,b,d,e) show equilibrium values for (unbounded) inner radius ah and thickness hh , as well as (bounded) circumferential σ θ h and axial σ zh Cauchy stress,
as functions of the stimulation-driver parameter ϕh = (coe − che )/coe . Panels (c) and (f) show the associated evolution of the homeostatic state in phase-type
planes: asymptotic thickening and dilation with moderate reduction in biaxial stress.
values for (a) ah , (b) hh , (d) σ θ h , and (e) σ zh as functions of the stimulation-driver ϕ h . Panels (c) and (f) show associ-
ated evolutions in phase planes. Following prior studies, which decrease either the mass fraction of elastin (Cyron et al.
2014; Zeinali-Davarani et al. 2011) or its elastic parameter (Valentín et al. 2013; Watton et al. 2004), we degraded che up to
ϕ h, max ≈ 0.8, namely ≈ 0.8coe in Eq. (119). Observe in this case that equilibrium values of inner radius and thickness, which
remain statically bounded initially, rapidly diverge when che approaches the value che = (1 − 0.775 )coe = 0.225coe , at which an
asymptotic growth response occurs. Consistent with Eq. (123), the problem becomes ill-posed during a quasi-static evo-
lution, with Jacobian determinants Do = 0.069 for ϕo = 0 and Dh → 0+ for ϕh → 0.775− . Conversely, equilibrium values of
intramural stresses decreased only slightly, remaining close to normal. These results are consistent with previous evolution
analyses and observations (cf. Valentín et al. 2013 and references therein) showing that irreversible damage to elastin pre-
vents an artery from maintaining its original homeostatic geometry and composition; with the passive stress contribution by
elastin diminished, the artery distends and collagen production increases in an attempt to compensate (Cyron et al. 2014).
A similar asymptotic growth response results for the pressure-driven case of Fig. 2, with che = coe constant, but with
Kτ = Kσ /5 and a high inner pressure of Ph ≈ 4.4Po , which could be reached only in cases of extreme adaptations such as
veins placed in the arterial system (Ramachandra et al. 2017). Note that this blow-up pressure, if any, depends on spe-
cific values of material parameters. Finally, the Jacobian determinants Dh computed for all the static cases studied evolved
continuously and were positive. According to Eq. (122), a potential change of sign of Dh (yh , ζ h ), occurring at a bounded
equilibrium state yh , could lead to singularities of some primary variables in the corresponding phase plane. At least in the
cases analyzed, with the present boundary conditions, constitutive relations and simulation-driver parameters, we observed
asymptotic responses and not limit points or bifurcations.
3.2. Mechanobiological stability of dynamic evolutions
3.2.1. Gain parameters Kσ and Kτ

The results in Section 3.1 describe how (static) equilibrium points of the system of Eqs. (107)–(109) evolve with respect
to sustained changes in external loads or model parameters (e.g., in Figs. 2 or 3). Ideally, in the absence of further per-
turbations, these solutions would remain statically equilibrated. We now employ Eqs. (107)–(109), including rate terms, to
describe how perturbed dynamic solutions behave near a previously equilibrated point under constant loads. Specifically,
consider perturbations of y(s) around the original homeostatic state yo with parameters in Table 1, that is, the initial equi-
librium point in Figs. 2 or 3. To show a precise correspondence between the integral-based formulation of Section 2.1 and
rate-based formulation of Section 2.2, we compute the eigenvalues of the Jacobian matrix in Eqs. (129)–(131) but also the
temporal evolution of the system using the integral-based formulation. For all cases analyzed, the transient (impulse-like)
perturbation consists of a rapid rise of inner pressure up to 1.5Po at s = 0+ days, which is maintained for 20 days and re-
turned back to the original homeostatic value Po at s = 20+ days. The dynamic stability character of y(s) around the original
state yo is then analyzed according to Eq. (125).
Fig. 4 shows five different dynamic responses associated with increasing absolute values of the gain parameters Kσ =
2Kτ > 0, which, based on Eq. (110), do not modify the original homeostatic state (because the terms within parentheses
vanish originally, i.e., ϒ o ≡ 1 in any case). For the first case, Kσ = 0.03, we obtain To = 0.149 > 0 and Do = 0.004 > 0, which
yields eigenvalues l2o/ko = 0.110 > 0 and l3o/ko = 0.040 > 0 in Eqs. (130) and (131), representing an asymptotically unsta-
ble (non-oscillatory) solution, or unstable source; see the phase plane in Fig. 4. The case Kσ = 0.12 yields To = 0.087 > 0
and Do = 0.018 > 0, with complex conjugates l2o and l3o given by Ro ≡ Re(l2o )/ko = 0.044 > 0 and Io ≡ Im(l2o )/ko = 0.125,
representing an asymptotically unstable (oscillatory) solution, or unstable spiral. The specific value Kσ = 0.2445 results in
a dynamic response that is neutrally stable around yo , with To = 0 and Do = 0.036 > 0, hence Ro = 0 and Io = 0.189, rep-
resented by a stable center. The oscillatory response becomes asymptotically stable for a further increase up to Kσ = 0.3,
with To = −0.038 < 0 and Do = 0.044 > 0, hence Ro = −0.019 < 0 and Io = 0.208, represented by a stable spiral in the phase
plane. Finally, for the higher value Kσ = 1.8, we obtain To = −1.075 < 0 and Do = 0.262 > 0, which yields real eigenval-
ues l2o/ko = −0.372 < 0 and l3o/ko = −0.703 < 0, representing an asymptotically stable (non-oscillatory) solution, or stable
sink in the phase plane. Indeed, based on the condition in Eq. (137), stable non-oscillatory responses were obtained for
Kσ = 2Kτ > 1.67.
Note the apparently low influence of neglected higher order terms in Eq. (128), especially for Kσ = 0.2445, which re-
mains neutrally stable, even for a moderate perturbation as the one introduced herein. Additional information regarding the
evolution of y(s) after removing the perturbation can be extracted from the eigenvalues computed from the linearized prob-
lem as rates of amplitude decay / grow (from their real part) or periods of oscillatory responses (from their imaginary part).
For example, for Kσ = 0.2445, Im(l2o ) = Ioko = 0.0189 days−1 , which yields an oscillatory period of 2π / Im(l2o ) ≈ 332 days,
consistent with the corresponding undamped response in Fig. 4 computed using the associated integral-based formulation.
Lastly, it can be shown that the same qualitative behavior is obtained for mean circumferential and axial stresses for the
different cases analyzed. For illustration, we show in Fig. 5 the dynamic evolution of σ θ and σ z for Kσ = 0.12 (first row,
unstable) and Kσ = 0.3 (second row, asymptotically stable). Hence, intramural stresses may eventually become unbounded
for dynamically unstable situations, whereas they remained bounded in the statically unstable cases simulated above, recall
Fig. 3. This fact highlights the importance of identifying what kind of instability develops in an arterial wall, if any.
3.2.2. Rate parameter ko

We comment here on the effect of the degradation rate parameter ko ≡ km c
o = ko > 0 on mechanobiological stability. First,
because this parameter is absent from Eqs. (110)–(112), different values of ko do not affect the mechanobiologically equi-
librated solutions given by Eq. (114); they merely influence when equilibration occurs. In case km c
o = ko , different ratios
kmo /k c yield different evolved homeostatic solutions, in general (Latorre and Humphrey 2018b). In the present case, with
o
km c
o /ko = ko /ko = 1, different values of ko yet affect the dynamics of the problem. Based on the eigenvalues in Eqs. (129)–
(131), or the original differential Eqs. (107)–(109), ko defines the time scale on which the system evolves but, because the
signs of (the real part of) the eigenvalues lio , i = 1, 2, 3, remain unchanged for different ko > 0, there is no change in the
mechanobiological stability character of the perturbed solutions. Albeit not shown, this was confirmed numerically using
our integral-based constrained mixture model with various values ko ; if the dynamic response after a transient perturbation
was asymptotically stable originally (e.g., fourth row in Fig. 4), an increase in ko made the response more stable, whereas if
it was unstable originally (e.g., second row in Fig. 4), an increase in ko made the response more unstable.
Substitution of Eq. (94) in Eq. (107) yields
ρ˙ R = ρ˙ R+ − ρ˙ R− = kKσ (ρR − ρoe )(σ − (Kτ /Kσ )τw ) (140)
which means that the net production of material (i.e., difference between total production ρ˙ R+
and total removal ρ˙ R− )
is pro-
portional to kKσ (cf., Cyron and Humphrey 2014 and Wu and Shadden 2016, wherein shear stress effects are not considered
and rate and gain parameters are combined into single non-dimensionless “gain” or “growth feedback” constants). Consid-
ering a reference case I and a case II with possible different production and removal rates, with Kτ /Kσ = const, an increased
production rate with a constant removal rate (i.e., kII = kI ) requires KσII > KσI in our formulation, such that the net produc-
tion ρ˙ RII > ρ˙ RI (for equal remaining variables). However, an increased removal rate (i.e., kII > kI ) for a constant production rate
requires KσII < (kI /kII )KσI < KσI , such that the net production ρ˙ RII < ρ˙ RI . Hence, because changes in ko do not modify the sta-
bility character of the system, while lower values of Kσ tend to destabilize it, we conclude that increased removal rates for
preserved production rates (i.e., kII > kI and KσII < (kI /kII )KσI ) tend to destabilize the system from a dynamic standpoint.
3.2.3. Material-to-prestress stiffness ratio cθ θ /(2σ θ )

Similarly to ko , the equilibrium components of the total material stiffness c(s) (cf. Eq. (29) for different constituents α ) are
absent from Eqs. (110)–(112), hence equilibrated stresses, rather than equilibrated values of stiffness, affect the mechanobio-
logically equilibrated solutions given by Eq. (114). Again based on Eqs. (132) and (133), however, the circumferential equilib-
rium stiffness cθ θ o may modify the dynamic stability character of perturbed solutions near the original homeostatic state. In
particular, the material-to-prestress stiffness ratio cθ θ o/(2σθneo ) = cθ θ o/(2σθ o − 2σθeo ) appears in both To and Do . Even though
the prestress σθneo = σθ o − σθeo only accounts for the stress of constituents that turnover, namely collagen and smooth muscle
cells, we analyze in this example the effect on dynamic stability of the (total) material-to-prestress stiffness ratio cθ θ o /(2σ θ o ).
The center row of panels in Fig. 4 showed a neutrally stable response around the original equilibrium state that follows
a 20-day impulse-like perturbation in inner pressure. For that case, Kσ = 2Kτ = 0.2445 and, from parameters in Table 1,
cθ θ o /(2σ θ o ) ≈ 2.5, which yielded Ro = Re(l2o )/ko = 0 and Io = Im(l2o )/ko = 0.189. We show in Fig. 6 results computed with
the same set of parameters except for the constants c1c and c2c for collagen, which were modified such that σ θ o remains
Fig. 4. Mechanobiologically unstable (Kσ = 0.03, “source”, and Kσ = 0.12, “spiral”; top two rows), neutral stable (Kσ = 0.2445, “center”; third row), and
asymptotic stable (Kσ = 0.3, “spiral”, and Kσ = 1.8, “sink”; bottom two rows) dynamic responses that follow a perturbation in pressure, consisting of a
rapid rise from Po up to 1.5Po at s = 0+ days, subsequently sustained for 20 days, and finally returned back to the original homeostatic value Po at s = 20+
days. For all cases, Kτ = Kσ /2.
constant while cθ θ o /(2σ θ o ) decreases to ≈ 1.5 (first row) or increases to ≈ 3.5 (second row). Clearly, a reduction in cir-
cumferential stiffness, for constant circumferential stress, which yielded Ro > 0, destabilizes the referential (neutral) stable
response, whereas an increase in cθ θ o /(2σ θ o ), which yielded Ro < 0, stabilizes it asymptotically.
Lastly, the condition in Eq. (137) predicts that the ratio cθ θ o /(2σ θ o ) should reach ≈ 29 (or higher) to yield a convergent
non-oscillatory response if the gain parameters remain as low as Kσ = 2Kτ = 0.2445. However, a more realistic, actually
attainable, minimum value for stable non-oscillatory evolutions cθ θ o /(2σ θ o ) ≈ 5 is predicted if the likely more physiological
(cf. Section 3.2.1) values Kσ = 2Kτ = 1 are considered. Changes in intrinsic material stiffness appear to be fundamental in
thoracic aortic aneurysms (Bellini et al. 2017), hence the importance of such considerations.
3.2.4. Active contraction of muscle

The prior results have focused on the passive response of an idealized arterial wall. We analyze in this example the
influence on mechanobiological stability of the active contribution to stress and stiffness given in Eqs. (64) and (69), along
with (65). The first row of panels in Fig. 7 show an asymptotically unstable (oscillatory) response near an associated original
homeostatic state computed with parameters in Table 1, except with Kσ = 2Kτ = 0.12 (cf. second row in Fig. 4) and an
additional active contribution to stress given by Sˆ = 40 kPa, CB = 0.8326, and CS = 0.5CB . The associated Jacobian matrix is
such that To = 0.068 > 0 and Do = 0.020 > 0, with complex eigenvalues l2o and l3o given by Ro = 0.034 > 0 and Io = 0.137.
Fig. 5. Dynamic stress response (σ θ and σ z ) for cases Kσ = 0.12 (cf. unstable spiral; second row, Fig. 4) and Kσ = 0.3 (cf. stable spiral; fourth row, Fig. 4).
Fig. 6. Dynamic stabilization afforded by an increased ratio of circumferential stiffness to material pre-stress cθ θ o /(2σ θ o ). Panels in first row show an
unstable oscillatory response with Kσ = 2Kτ = 0.2445 and cθ θ o /(2σ θ o ) ≈ 1.5 (cf. third row, Fig. 4, wherein Kσ = 2Kτ = 0.2445 and cθ θ o /(2σ θ o ) ≈ 2.5). Panels
in second row show results for cθ θ o /(2σ θ o ) ≈ 3.5, which stabilizes asymptotically the prior neutrally stable response.
We analyze three different modifications of the active response given by the different parameters Sˆ, CB , and CS . A fold
increase in the second Piola-Kirchhoff stress-like active tone Sˆ yields the same fold increase in both equilibrated active stress,
which from Eq. (64), with λθ o = 1, τw |o = 0, and Co = CB , reads
σθact
o = φo S ( 1 − e
−CB
),
2
mˆ
(141)
−CB2
and equilibrated active-like stiffness, which from Eq. (69), with τwo/τwo = 1 and dSˆθact /dC |o = 2SˆCB e , reads
cθactθ o = 6φ mˆ
o SCSCB e
−CB2
. (142)
A fold increase in the basal ratio CB yields a nonlinear increase in stress, which saturates for high increments of CB
σθact
o → φo S > 0 for CB ↑,
mˆ
(143)
and a nonlinear decrease in stiffness, which tends to vanish
cθactθ o → 0+ for CB ↑. (144)
Finally, a fold increase in the vasoactive parameter CS yields no change in stress σθact ,
but the same fold change in stiffness
o
cθact
θ o.
According to Eqs. (132) and (133), along with Eqs. (135) and (136), different effects on σθact o
and cθact
θ o, and hence on
the combined stiffness Cθ θ o = 2σθ o + cθ θ o, lead to different dynamic responses for a given initial perturbation. We show in
act act act
Fig. 7, second to fourth rows, computed dynamic responses for respective 10-fold increases in Sˆ, CB , and CS with respect
to the reference case (first row). Interestingly, higher values of Sˆ (second row), which increase σ act and cact proportionally, θo θθo
and CS (fourth row), which increase cθact
θ o, stabilize (Ro = −0.042 and Ro = −0.027, respectively) the reference response (Ro =
Fig. 7. Dynamic stabilization by the active response of smooth muscle cells. Panels in first row show an unstable response with passive (Table 1, except
for Kσ = 2Kτ = 0.12; cf. second row, Fig. 4) and additional active (Sˆ = 40 kPa, CB = 0.8326, CS = 0.5CB ) contributions to stress. A 10-fold increase in either
Sˆ (second row) or CS (fourth row) stabilize the prior response (first row). A 10-fold increase in CB (third row) has little effect over the prior response, even
augmenting the instability.
Table 1
Representative baseline model parameters for a mouse descending thoracic aorta, adapted (homogenized
through the thickness) from Latorre and Humphrey (2018a). Superscripts e, m, c denote elastin, smooth
muscle, and collagen, with superscripts/subscripts r, θ , z, d denoting radial, circumferential, axial, and
symmetric diagonal directions. Subscript o denotes original homeostatic values. Subscripts σ and τ de-
note intramural and wall shear stress related parameters, respectively.
Artery mass density ρ 1050 kg/m3

Inner radius, wall thickness ao , ho [0.6468, 0.0402] mm
Mass fractions φoe , φom , φoc [0.30, 0.35, 0.35]
Collagen fractions βθ , βz, βd [0.068, 0.381, 0.551]
Diagonal collagen orientation α0 45.36o
Elastin parameter ce 114.5 kPa
Smooth Muscle parameters c1m , c2m [401.0 kPa, 0.012]
Collagen parameters c1c , c2c [411.2 kPa, 5.5]
Deposition stretches Ger , Geθ , Gez [1/(1.9 · 1.6), 1.9, 1.6]
Gm θ = Gθ , Gz , Gd
c c c
Deposition stretches [1.071, 1.193, 1.192]
Mass production gains Kσ , Kτ [1.0, 0.5]
Mass removal rate ko 1/10 day−1
0.034), whereas an equal fold increase in CB (third row), with associated saturated stress σθact
o
and vanishing stiffness cθact
θ o,
yields a slightly more unstable response (Ro = 0.040) than the referential one. Albeit not shown, intramural stresses for each
case followed respective stable (second, fourth rows) or unstable (first, third rows) dynamic evolutions. Lastly, among the
many possible combinations for these parameters, Eq. (137) predicts stable non-oscillatory G&R responses for a combined
7.07- or higher-fold change in both Sˆ and CS , with CB unchanged. The importance of active stress generation has long been
known in general arterial G&R (Valentín et al. 2009) and recently was revealed in aortic dissection (Ferruzzi et al. 2016).
4. Discussion
Years ago, Waxman (1981) considered axial (buckling) instabilities in arteries as a function of prior biological growth.
Although inappropriately based on linear elasticity, this paper highlighted the need to study growth and related issues of
stability within a framework of continuum mechanics. Soon thereafter, Skalak et al. (1982) emphasized the need to study
growth in terms of finite displacements, that is, nonlinear continuum mechanics. They also noted the possibility of unstable
growth if one does not consider appropriate interdependencies in rates of allometric growth of individual structural ele-
ments. No formal stability analysis was offered, however. Somewhat surprisingly, growth mechanics did not advance much
further until mid-1990s, with the introduction of a theory of finite kinematic growth. Among others who adopted this ap-
proach, Taber (1998) suggested that evolution laws for arterial adaptations to altered hemodynamics must include negative
feedback, in terms of homeostatic target values, to yield stable responses. Further increasing interest in modeling G&R was
the introduction of a constrained mixture approach that enabled one to consider different rates of turnover of different
constituents (Humphrey and Rajagopal 2002).
There now exist many different models of soft tissue G&R based on mixture theory. For example, Klisch et al. (2003) used
a mixture model to describe independent contributions of proteoglycans and collagen to the growth of car-
tilage. Lemon et al. (2006) used a porous flow mixture theory to study growth of engineered tissues ex
vivo. Cristini et al. (2009) used a multi-phase mixture model to study the growth of avascular solid tumors.
Narayanan et al. (2009) presented a general mixture-based model of growth that coupled mass transport and tissue me-
chanics. Haider et al. (2011) used a mixture model to study matrix production in a cell-seeded tissue engineered scaf-
fold for cartilage. Cowin and Cardoso (2012) proposed a general mixture-based poroelastic model of interstitial growth.
Ateshian et al. (2014) modeled interstitial tissue growth by considering both the solid mechanics and biochemical reac-
tions. Soares and Sacks (2016) used a triphasic constrained mixture model to describe engineered tissue formation under
in vitro dynamic mechanical loading. Vernerey (2016) used a constrained mixture model of interstitial growth in polymeric
scaffolds for tissue engineering. Truster and Masud (2017) similarly used a mixture theory to study the infiltration of cells
and neotissue formation within degrading polymeric scaffolds used for tissue engineering. Additional discussion of these
and other mixture approaches can be found in Ambrosi et al. (2011) and Ateshian and Humphrey (2012). Note, too, that
Watton et al. (2004) and Baek et al. (2006) used mixture-based computational models to study aneurysmal G&R. Whereas
the former reported an unstable / unbounded enlargement of these lesions, the latter showed that multiple parameters (e.g.,
rates of tissue production and preferential alignment of the new tissue) can stabilize the enlargement, at least when produc-
tion rates depend on differences in stress from homeostatic targets, consistent with the suggestion of Taber (1998). Again,
however, these works did not consider formal stability analyses. In contrast, Ben Amar and Goriely (2005) noted that growth
can alter both the geometry of and residual stress field within a tissue, each of which can affect mechanical stability under
near static loading. Specifically, these authors contrasted cases wherein prior growth could either stabilize (via wall thick-
ening or tensile residual stresses) or destabilize (via thinning or compressive residual stresses) a prototypical thick-walled,
neoHookean spherical shell model of a tissue. Revisiting the study of Waxman (1981), Goriely and Vandiver (2010) used
similar ideas to study axial buckling of arteries subjected to increasing pressures.
More recent studies have instead focused on the stability of the G&R process itself (i.e., mechanobiological stability)
rather than the mechanical stability that results from prior G&R. Erlich et al. (2018) studied the possible stability of the
growth of layered tubular structures that exhibit an isotropic, materially uniform behavior; they used the concept of kine-
matic growth rather than a mixture approach. Additionally, Satha et al. (2014), Cyron et al. (2014) and Wu and Shad-
den (2016) used different mixture approaches to study the stability of G&R in arteries and arrived at consistent conclusions.
Increased tissue stiffness and higher rates of tissue production tend to stabilize G&R processes under constant pressures
while increased rates of degradation can destabilize the G&R. Hence, there is an acute need to consider parameter values
carefully, both to identify optimal G&R capacity and to determine whether a process will or will not be mechanobiologically
stable.
In this paper, we sought both to develop a general framework and to illustrate consequences of G&R stabilities by per-
forming a systematic, formal analysis for idealized arteries while yet accounting for complexities such as different nonlinear
material properties and rates of turnover for different structurally significant constituents and including intramural biaxial
and wall shear stresses as stimuli for mass production as well as active and passive contributions of smooth muscle. Toward
this end, we first derived a new rate-based form for a constrained mixture that is equivalent to the traditional heredity
integral-based form but facilitates linearized stability analyses about appropriate mechanobiological equilibria. Numerical
simulations confirmed this equivalency, at least for the canonical problems considered. This rate-form revealed, among other
findings, the appropriateness of the Truesdell stress rate in constrained mixture G&R formulations and the natural separation
of rates of change of stress into elastic and inelastic parts, the latter in terms of “growth” and “remodeling” aspects. We next
derived a system of nonlinear ordinary differential equations for G&R of a cylindrical artery that admitted both an analysis
of critical points and an eigenvalue analysis of respective linearized systems. The former led to the concept of mechanobi-
ological static stability of mechanobiologically equilibrated states, which allows one to analyze how soft tissues adapt to
sustained changes in external loads or material properties, with G&R time s conveniently eliminated from the analysis; the
latter led to the concept of mechanobiological dynamic stability of perturbed solutions around previously equilibrated states.
For the arterial model considered, these analyses delineated two different types of possible instabilities, namely, asymptotic
growth of static equilibria and asymptotic growth of dynamic responses around (bounded) equilibrated states.
Regarding the novel concept of mechanobiological static stability of G&R states, our findings are consistent with prior
studies that showed a destabilizing effect of elastin degradation (cf. Valentín et al. 2013; Watton et al. 2004; Zeinali-Davarani
et al. 2011). Yet, even though this G&R problem in the aorta, leading to aneurysm growth, is frequently addressed from a
time-dependent perspective, we confirmed that its ultimate cause may be a mechanobiological instability of the evolving
equilibrium state (Cyron et al. 2014), which we directly addressed with a static approach. Analyses of this type driven by
other external loads or material parameters could lead to other types of static instabilities, such as limit points or bifurca-
tions.
Additional findings related to the mechanobiological stability of transiently perturbed evolutions around equilibrium
states were consistent with other studies that showed the stabilizing effects of increased material stiffness and tissue pro-
duction rates and destabilizing effects of increased removal rates (cf. Cyron et al. 2014; Satha et al. 2014; Wu and Shadden
2016). At least for the constitutive relations and model parameters used, we additionally showed that gain parameters for
mass production that are associated with both intramural biaxial and wall shear stresses play important roles in the eigen-
value analysis and that an increasing active contribution to stress was stabilizing. Consistent with the type and sign of
the eigenvalues of the Jacobian matrix, determined analytically for the original homeostatic state, numerical simulations
further revealed that unstable, neutrally stable, or asymptotically stable results could arise mathematically in response to
transient perturbations depending on specific model assumptions or values of the parameters, hence extending the analysis
of Satha et al. (2014). In this regard, note that we focused on the stability of an idealized and isolated growing artery from
a constitutive point of view, disregarding other factors that could damp oscillations, including external perivascular support
or additional intrinsic dissipation. Nevertheless, as we illustrated in examples above, mathematical analyses of this type can
serve to identify conditions that ensure a physiologically reasonable dynamics predicted by the G&R constitutive model.
Our theory, based originally on fully coupled nonlinear evolutions of mass and stress (Fung 1995), thus builds on the
mechanobiological stability theory of Cyron and Humphrey (2014), where mechanical stability (against displacement per-
turbations) and mechanobiological stability (against mass perturbations) were analyzed incrementally based on a theory of
small on large (Baek et al. 2007a) extended to G&R by distinguishing elastic and inelastic deformations. Mechanobiolog-
ically stable systems considered therein, with sustained and transient responses analyzed together, were always neutrally
(Lyapunov) stable under small, residual perturbations with respect to the original state, leading to associated definitions
of mechanobiological adaptivities. Using a linearized stress-strain relation with respect to a homeostatic state of the vessel
wall, among other assumptions, Wu and Shadden (2016) also found a neutrally stabilizing condition, or degeneration, when
the dynamics of displacement and mass variables were decoupled from the dynamics of an additional generalized stiffness
variable. Further use of an extended system led to the main stability conclusions that displacements and mass were neutrally
stable whereas stress and stiffness were asymptotically stable. In contrast, our formulation, based on nonlinear constitutive
relations for multiple constituents that may turnover or not, did not lead, by default, to a degenerate system of this kind;
rather, neutral stability in the sense of Lyapunov around evolved critical points was obtained only for particular cases. In
general, mechanobiological dynamic stability following the nonlinear equations considered herein exhibited an asymptotic
character, either oscillatory or not, under moderate perturbations. Because of the lack of empirical evidence of oscillatory
behaviors in vivo, these findings can aid further in refining values of parameters that are physiologically meaningful and yet
revealing.
In conclusion, it is becoming increasingly evident that mechanical homeostasis is fundamental to healthy tissue structure
and function and conversely that compromised or lost homeostasis underlies many cases of disease (Humphrey et al. 2014).
Homeostasis necessarily implies negative feedback loops that govern G&R processes at the tissue level, with target values
of appropriate mechanical metrics. It has recently been suggested that the converse, that is positive feedback loops, implies
disease progression, leading to a type of biological instability (Schwartz et al. 2018). We suggest that mechanobiological sta-
bility analyses, such as those performed herein, promise to provide increasing insight into general processes that promote
health versus disease progression. There is, therefore, a pressing need to continue to broaden and advance our constitutive
relations for G&R, building on new biological findings as they become available while continuing to capture fundamen-
tal features of soft tissues, including the different material properties, rates of turnover, and natural configurations of the
different structural constituents.
Conflict of interest
The authors declare that they have no conflict of interest.
Acknowledgments
This work was supported, in part, by grants from the US NIH, namely, R01 HL105297 (to C.A. Figueroa and J.D.
Humphrey), R01 HL128602 (to J.D. Humphrey, C.K. Breuer, and Y. Wang), P01 HL134605 (to G. Tellides and J.D. Humphrey
via a PPG to D. Rifkin), and U01 HL142518 (to J.D. Humphrey and G.E. Karniadakis).
Appendix A. Constitutive tangent moduli
Elastin
φoe 2
cejk (s ) = λ (s )λ2k(s )Gej2 Gek2Cêjk (s ) j, k = {r, θ , z} (A.1)
J(s ) j
Smooth muscle
λ4θ (s ) m
s
ρ m(τ )Gm
θ Cθ θ(s, τ )
4 ˆm
cθmθ (s ) = k (τ )ϒ m(τ )qm(s, τ ) R dτ (A.2)
J ( s )ρ λ4θ(τ )
−∞
Collagen
λ4 (s ) c
s
ρ c (τ )Gc4
θ Cθ θ(s, τ )
ˆc
cθ θ ( s ) = θ
c
k (τ )ϒ c(τ )qc(s, τ ) Rθ dτ
J ( s )ρ λ4θ(τ )
−∞
λθ (s ) c
s
4
ρ c (τ )Gc4 Cˆc s, τ )
d dd(
k (τ )ϒ c(τ )qc(s, τ ) Rd sin αd dτ
4
+ (A.3)
J ( s )ρ λθ(τ )
4
−∞
s
λ4z (s ) ρRz
c
(τ )Gc4
z Czz(s, τ )
ˆc
c
czz (s ) = kc(τ )ϒ c(τ )qc(s, τ ) dτ
J ( s )ρ λz(τ )
4
−∞
s
λ4z (s ) ρRd
c
(τ )Gc4 Cˆc s, τ )
d dd(
+ kc(τ )ϒ c(τ )qc(s, τ ) cos4 αd dτ (A.4)
J ( s )ρ λ4z(τ )
−∞
λ2θ (s )λ2z (s ) c
s
ρ c (τ )Gc4Cˆc (s, τ ) 2
cθc z (s ) = k (τ )ϒ c(τ )qc(s, τ ) Rd 2 d 2dd sin αd cos2 αd dτ (A.5)
J(s )ρ λθ(τ )λz(τ )
−∞
Active contribution
dSˆθact (C (s ) ) τw (s )
cθactθ (s ) = 3φ m(s )λ2θ (s )CS (A.6)
dC (s ) τwo
Mechanobiologically equilibrated values

e,m,c,act
c c2
σθ o = σθαo = φoe Sˆθe Geθ2 + φom Sˆθm Gm
θ + φo βθ Sθ Gθ + βd Sd Gd sin αd + σθ o
2 c ˆ ˆc c2 2 act
(A.7)
α
σθeo = φoe σˆ θe = φoe Sˆθe Geθ2 , σˆ zoe = Sˆze Gez2 (A.8)

e,c
e,c

σzo = σzoα = φoα σˆ zα = φoe Sˆze Gez2 + φoc βz Sˆzc Gc2
z + βd Sd Gd cos αd
ˆc c2 2
(A.9)
α α

σθact
o = φo S 1 − e
−CB2
cθactθ o = 6φom SˆCSCB e−CB
2
mˆ
, (A.10)

e,m,c,act
act
cθαθ o = 0 + φomCˆθmθ Gm
θ + φo θ + βd Cdd Gd sin αd + cθ θ o
βθ Cˆθc θ Gc4
4 c ˆc c4 4
cθ θ o = (A.11)
α

e,c
cθαzo = 0 + φoc βdCˆdd d sin αd cos αd = φo βd Cdd Gd sin αd cos αd
c 2 ˆc c4 2
cθ zo = Gc4 2 c 2
(A.12)
α
Appendix B. Jacobian matrix derivatives
Auxiliary results
∂ (τw /τwo )
3 ρσθ o 1/2 ρσθ o 3 τw 1 o 3
=− =− −→ − (B.1)
∂ρR 2 ρR σ θ ρR σ θ
2 2 τ ρ
wo R 2 ρ
∂ (τw /τwo )
3 ρσθ o 1/2 ρσθ o 3 τw 1 o 3
=− =− −→ − (B.2)
∂σθ 2 ρR σ θ ρR σθ2 2 τwo σθ 2 σθ o
∂ϒ ∂ (τw /τwo ) o 3Kτ K̄τ

= −Kτ −→ = (B.3)
∂ρR ∂ρR 2ρ ρ
∂ϒ ∂ ((σθ + σz )/σIo ) ∂ (τw /τwo ) o Kσ 3Kτ K̄σ + K̄τ

= Kσ − Kτ −→ + = (B.4)
∂σθ ∂σθ ∂σθ σIo 2σθ o σθ o
∂ϒ ∂ ((σθ + σz )/σIo ) o Kσ K̄σ

= Kσ −→ = (B.5)
∂σz ∂σz σIo σθ o

∂ Sˆθact dSˆθact dC ∂ (τw /τwo ) dSˆθact 3CS
cact
= = = θ θ om (B.6)
∂ρR o dC d (τw /τwo ) ∂ρR o
dC
o
2ρ 2ρφo

∂ Sˆθact dSˆθact dC ∂ (τw /τwo ) dSˆθact 3CS
cθactθ o
= = = (B.7)
∂σθ o dC d (τw /τwo ) ∂σθ o
dC
o
2 σθ o 2σθ oφom
Jacobian derivatives at original homeostatic state o

1 ∂ ρ˙ R ρ − ρoe
= K̄τ (B.8)
ko ∂ρR o ρ

1 ∂ ρ˙ R ρ − ρoe
= K̄σ + K̄τ (B.9)
ko ∂σθ σ θ o
o

1 ∂ ρ˙ R ρ − ρoe
= K̄ (B.10)
ko ∂σz σθ o σ
o
ne
ρ ∂ σ˙ θ σθ o ρ − ρoe σ ne ρoe C act
− = 2 + K̄τ + 2 θ o + θθo (B.11)
koσθ o ∂ρR cθ θ o ρ c θ θ o ρ − ρo e
cθ θ o
o
ne
1 ∂ σ˙ θ σθ o ρ − ρoe σθneo Cθactθ o
− = 2 + K̄ + K̄ − 2 + (B.12)
ko ∂σθ
σ τ
o
cθ θ o ρ cθ θ o cθ θ o
ne
1 ∂ σ˙ θ σθ o ρ − ρoe
− = 2 + K̄σ (B.13)
ko ∂σz cθ θ o ρ
o

ρ ∂ σ˙ z cθ zo σθneo σzoe ρoe σzo cθ zo σθneo ρoe
− = + − K̄ +
koσθ o ∂ρR
τ
c θ θ o σθ o σθ o ρ σθ o cθ θ o σθ o ρ − ρoe
o
e
σzo ρoe σzo ρ cθ zo Cθact
θo
+ − + (B.14)
σθ o ρ σθ o ρ − ρoe 2σθ o cθ θ o

1 ∂ σ˙ z cθ zo σθneo σzoe ρoe σzo c 2σθneo + Cθactθ o
− = + − K̄σ + K̄τ + θ zo (B.15)
ko ∂σθ c θ θ o σθ o σθ o ρ σθ o 2 σθ o cθ θ o
o

1 ∂ σ˙ z cθ zo σθneo σ e ρ e σzo
− =1+ + zo − o K̄ (B.16)
ko ∂σz
o
c θ θ o σθ o σθ o ρ σθ o σ
References
Ambrosi, D., Ateshian, G.A., Arruda, E.M., Cowin, S.C., Dumais, J., Goriely, A., Holzapfel, G.A., Humphrey, J.D., Kemkemer, R., Kuhl, E., Olberding, J.E., Taber, L.A.,
Garikipati, K., 2011. Perspectives on biological growth and remodeling. J. Mech. Phys. Solids 59 (4), 863–883.
Ateshian, G.A., Humphrey, J.D., 2012. Continuum mixture models of biological growth and remodeling: past successes and future opportunities. Annu. Rev.
Biomed. Eng. 14, 97–111.
Ateshian, G.A., Nims, R.J., Maas, S., Weiss, J.A., 2014. Computational modeling of chemical reactions and interstitial growth and remodeling involving charged
solutes and solid-bound molecules. Biomech. Model. Mechanobiol. 13 (5), 1105–1120.
Baek, S., Gleason, R.L., Rajagopal, K.R., Humphrey, J.D., 2007a. Theory of small on large: potential utility in computations of fluid–solid interactions in
arteries. Comput. Methods Appl. Mech. Eng. 196 (31–32), 3070–3078.
Baek, S., Rajagopal, K.R., Humphrey, J.D., 2006. A theoretical model of enlarging intracranial fusiform aneurysms. J. Biomech. Eng. 128 (1), 142–149.
Baek, S., Valentín, A., Humphrey, J.D., 2007b. Biochemomechanics of cerebral vasospasm and its resolution: II. constitutive relations and model simulations.
Ann. Biomed. Eng. 35 (9), 1498.
Bellini, C., Bersi, M.R., Caulk, A.W., Ferruzzi, J., Milewicz, D.M., Ramirez, F., Rifkin, D.B., Tellides, G., Yanagisawa, H., Humphrey, J.D., 2017. Comparison of 10
murine models reveals a distinct biomechanical phenotype in thoracic aortic aneurysms. J. R. Soc. Interf. 14 (130), 20161036.
Bellini, C., Ferruzzi, J., Roccabianca, S., Di Martino, E.S., Humphrey, J.D., 2014. A microstructurally motivated model of arterial wall mechanics with
mechanobiological implications. Ann. Biomed. Eng. 42 (3), 488–502.
Ben Amar, M., Goriely, A., 2005. Growth and instability in elastic tissues. J. Mech. Phys. Solids 53 (10), 2284–2319.
Bersi, M.R., Bellini, C., Wu, J., KRC, M., Harrison, D.G., Humphrey, J.D., 2016. Excessive adventitial remodeling leads to early aortic maladaptation in an-
giotensin-induced hypertension. Hypertension 67, 890–896.
Boyce, W.E., DiPrima, R.C., 2012. Elementary differential equations and boundary value problems, Tenth edition. JohnWiley & Sons, New Jersey.
Cowin, S.C., Cardoso, L., 2012. Mixture theory-based poroelasticity as a model of interstitial tissue growth. Mech. Mater. 44, 47–57.
Cristini, V., Li, X., Lowengrub, J.S., Wise, S.M., 2009. Nonlinear simulations of solid tumor growth using a mixture model: invasion and branching. J. Math.
Biol. 58 (4–5), 723.
Cyron, C.J., Aydin, R.C., Humphrey, J.D., 2016. A homogenized constrained mixture (and mechanical analog) model for growth and remodeling of soft tissue.
Biomech. Model. Mechanobiol. 15 (6), 1389–1403.
Cyron, C.J., Humphrey, J.D., 2014. Vascular homeostasis and the concept of mechanobiological stability. Int. J. Eng. Sci. 85, 203–223.
Cyron, C.J., Wilson, J.S., Humphrey, J.D., 2014. Mechanobiological stability: a new paradigm to understand the enlargement of aneurysms? J. R. Soc. Interf.
11 (100), 20140680.
Erlich, A., Moulton, D.E., Goriely, A., 2018. Are homeostatic states stable? dynamical stability in morphoelasticity. Bull. Math. Biol. doi:10.1007/
s11538- 018- 0502- 7.
Ferruzzi, J., Murtada, S.I., Li, G., Jiao, Y., Uman, S., MYL, T., Tellides, G., Humphrey, J.D., 2016. Pharmacologically improved contractility protects against aortic
dissection in mice with disrupted transforming growth factor-β signaling despite compromised extracellular matrix properties. Arterioscler. Thromb.
Vasc. Biol. 36, 919–927.
Fung, Y.C., 1995. Stress, Strain, Growth, and Remodeling of Living Organisms. In: Casey, J., Crochet, M.J. (Eds.), Theoretical, Experimental, and Numerical
Contributions to the Mechanics of Fluids and Solids, Birkhäuser, Basel, pp. 469–482.
Gleason, R.L., Wilson, E., Humphrey, J.D., 2007. Biaxial biomechanical adaptations of mouse carotid arteries cultured at altered axial extension. J. Biomech.
40 (4), 766–776.
Goriely, A., Vandiver, R., 2010. On the mechanical stability of growing arteries. IMA J. Appl. Math. 75 (4), 549–570.
Haider, M.A., Olander, J.E., Arnold, R.F., Marous, D.R., McLamb, A.J., Thompson, K.C., Woodruff, W.R., Haugh, J.M., 2011. A phenomenological mixture model
for biosynthesis and linking of cartilage extracellular matrix in scaffolds seeded with chondrocytes. Biomech. Model. Mechanobiol. 10 (6), 915–924.
Hairer, E., Nørsett, S.P., Wanner, G., 1993. Solving Ordinary Differential Equations: I. Nonstiff Problems, 2nd Edn. Series in Computational Mathematics,
Vol. 8. Springer Verlag.
Haslach, H.W., Humphrey, J.D., 2004. Dynamics of biological soft tissue and rubber: internally pressurized spherical membranes surrounded by a fluid. Int.
J. Non Linear Mech. 39 (3), 399–420.
Holzapfel, G.A., 1996. On large strain viscoelasticity: continuum formulation and finite element applications to elastomeric structures. Int. J. Numer. Methods
Eng. 39 (22), 3903–3926.
Holzapfel, G.A., 20 0 0. Nonlinear Solid Mechanics. A Continuum Approach for Engineering. John Wiley & Sons, Chichester.
Humphrey, J.D., 2002. Cardiovascular Solid Mechanics: Cells. Tissues and Organs. Springer-Verlag, New York.
Humphrey, J.D., Dufresne, E.R., Schwartz, M.A., 2014. Mechanotransduction and extracellular matrix homeostasis. Nat. Rev. Mol. Cell Biol. 15 (12), 802–812.
Humphrey, J.D., Eberth, J.F., Dye, W.W., Gleason, R.L., 2009. Fundamental role of axial stress in compensatory adaptations by arteries. J. Biomech. 42 (1),
1–8.
Humphrey, J.D., Rajagopal, K.R., 2002. A constrained mixture model for growth and remodeling of soft tissues. Math. Models Methods Appl. Sci. 12 (03),
407–430.
Hynes, R.O., Naba, A., 2012. Overview of the matrisome: an inventory of extracellular matrix constituents and functions. Cold Spring Harb. Perspect. Biol. 4
(1). A004903.
Klisch, S.M., Chen, S.S., Sah, R.L., Hoger, A., 2003. A growth mixture theory for cartilage with application to growth-related experiments on cartilage explants.
J. Biomech. Eng. 125 (2), 169–179.
Latorre, M., Humphrey, J.D., 2018a. Critical roles of time-scales in soft tissue growth and remodeling. APL Bioeng. 2 (2), 026108.
Latorre, M., Humphrey, J.D., 2018b. A mechanobiologically equilibrated constrained mixture model for growth and remodeling of soft tissues. ZAMM-J. Appl.
Math. Mech. 98, 2048–2071.
Latorre, M., Humphrey, J.D., 2018. Modeling mechano-driven and immuno-mediated aortic maladaptation in hypertension. Biomech. Model. Mechanobiol.
17 (5), 1497–1511.
Latorre, M., Montáns, F.J., 2015. Anisotropic finite strain viscoelasticity based on the Sidoroff multiplicative decomposition and logarithmic strains. Comput.
Mech. 56 (3), 503–531.
Lemon, G., King, J.R., Byrne, H.M., Jensen, O.E., Shakesheff, K.M., 2006. Mathematical modelling of engineered tissue growth using a multiphase porous flow
mixture theory. J. Math. Biol. 52 (5), 571–594.
Lyapunov, A.M., 1882. Problème général de la stabilité du mouvement. Annals of Mathematics Studies, Vol. 17. Princeton University Press. Reprint 1947.
Miller, K.S., Lee, Y.U., Naito, Y., Breuer, C.K., Humphrey, J.D., 2014. Computational model of the in vivo development of a tissue engineered vein from an
implanted polymeric construct. J. Biomech. 47 (9), 2080–2087.
Murtada, S.I., Humphrey, J.D., Holzapfel, G.A., 2017. Multiscale and multiaxial mechanics of vascular smooth muscle. Biophys. J. 113 (3), 714–727.
Narayanan, H., Arruda, E.M., Grosh, K., Garikipati, K., 2009. The micromechanics of fluid–solid interactions during growth in porous soft biological tissue.
Biomech. Model. Mechanobiol. 8 (3), 167.
Ramachandra, A.B., Humphrey, J.D., Marsden, A.L., 2017. Gradual loading ameliorates maladaptation in computational simulations of vein graft growth and
remodelling. J. R. Soc. Interf. 14 (130), 20160995.
Rouche, N., Habets, P., Laloy, M., 1977. Stability Theory by Liapunov’s Direct Method. In: Applied Mathemathical Sciences, Vol. 22. Springer Verlag.
Satha, G., Lindström, S.B., Klarbring, A., 2014. A goal function approach to remodeling of arteries uncovers mechanisms for growth instability. Biomech.
Model Mechanobiol. 13 (6), 1243–1259.
Schwartz, M.A., Vestweber, D., Simons, M., 2018. A unifying concept in vascular health and disease. Science 360 (6386), 270–271.
Sidoroff, F., 1974. Un modèle viscoélastique non linéaire avec configuration intermédiaire. Journal de Mécanique 13 (4), 679–713.
Simó, J.C., 1987. On a fully three-dimensional finite-strain viscoelastic damage model: formulation and computational aspects. Comput. Methods Appl. Mech.
Eng. 60 (2), 153–173.
Simó, J.C., Pister, K.S., 1984. Remarks on rate constitutive equations for finite deformation problems: computational implications. Comput. Methods Appl.
Mech. Eng. 46 (2), 201–215.
Skalak, R., Dasgupta, G., Moss, M., Otten, E., Dullemeijer, P., Vilmann, H., 1982. Analytical description of growth. J. Theor. Biol. 94 (3), 555–577.
Soares, J.S., Sacks, M.S., 2016. A triphasic constrained mixture model of engineered tissue formation under in vitro dynamic mechanical conditioning.
Biomech. Model Mechanobiol. 15 (2), 293–316.
Taber, L., 1998. A model for aortic growth based on fluid shear and fiber stresses. J. Biomech. Eng. 120 (3), 348–354.
Truster, T.J., Masud, A., 2017. A unified mixture formulation for density and volumetric growth of multi-constituent solids in tissue engineering. Comput.
Methods Appl. Mech. Eng. 314, 222–268.
Valentín, A., Cardamone, L., Baek, S., Humphrey, J.D., 2009. Complementary vasoactivity and matrix remodelling in arterial adaptations to altered flow and
pressure. J. R. Soc. Interf. 6 (32), 293–306.
Valentín, A., Humphrey, J.D., 2009. Evaluation of fundamental hypotheses underlying constrained mixture models of arterial growth and remodelling. philo-
sophical transactions of the royal society of london a: mathematical. Phys. Eng. Sci. 367 (1902), 3585–3606.
Valentín, A., Humphrey, J.D., Holzapfel, G.A., 2013. A finite element-based constrained mixture implementation for arterial growth, remodeling, and adap-
tation: theory and numerical verification. Int. J. Numer. Method. Biomed. Eng. 29 (8), 822–849.
Vernerey, F.J., 2016. A mixture approach to investigate interstitial growth in engineering scaffolds. Biomech. Model Mechanobiol. 15 (2), 259–278.
Wagenseil, J.E., Mecham, R.P., 2009. Vascular extracellular matrix and arterial mechanics. Physiol. Rev. 89 (3), 957–989.
Watton, P.N., Hill, N.A., Heil, M., 2004. A mathematical model for the growth of the abdominal aortic aneurysm. Biomech. Model. Mechanobiol. 3 (2),
98–113.
Waxman, A.M., 1981. A continuum approach to blood vessel growth: axisymmetric elastic structures. J. Theor. Biol. 91 (2), 273–301.
Wu, J., Shadden, S.C., 2016. Stability analysis of a continuum-based constrained mixture model for vascular growth and remodeling. Biomech. Model
Mechanobiol. 15 (6), 1669–1684.
Zeinali-Davarani, S., Sheidaei, A., Baek, S., 2011. A finite element model of stress-mediated vascular adaptation: application to abdominal aortic aneurysms.
Comput. Methods Biomech. Biomed. Eng. 14 (9), 803–817.

1 s2.0 S0022509618307439 Main

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

1 s2.0 S0022509618307439 Main

Uploaded by

Copyright:

Available Formats

Journal of the Mechanics and Physics of Solids 125 (2019) 298–325

Contents lists available at ScienceDirect

Journal of the Mechanics and Physics of Solids

Mechanobiological stability of biological soft tissues

2.1. Constrained mixture framework: Integral form

where Jnα(τ ) (s ) = det Fα

mαR(τ ) = kα(τ )ρRα(τ )ϒ α(τ ) (11)

we consider mechanobiological stability of previously equilibrated mechanobiological states, mα

2.2. Constrained mixture framework: Rate form

ρ˙ Rα(s ) = kα(s )ρRα(s )(ϒ α(s ) − 1 ) (13)

which involves the rates F˙ α s , J˙α s and ∂ Sˆ α (Cα

∂ Sˆ α (Cαn(τ )(s ) ) ∂ 2Wˆ α (Cαn(τ )(s ) )

σ α (s ) = kα (s ) ϒ α (s )σ αdep (s ) − σ α (s ) + cα (s ) : d(s ) (31)

Sαdep (s ) := J (s )F−1 (s )σ αdep (s )F−T (s ) (34)

2.3. A constrained mixture model for arterial G&R in rate form

reads for each (principal) component

σr (s ) = σre (s ) − p(s ), (42)

2.3.2. Smooth muscle

directionally (Appendix A).

2.3.4. Active stress

where C˙ reads, from Eqs. (65) and (66),

and the circumferential stretch for active stresses be approximated by

σ˙ θact (s ) = km (s ) ϒ m (s )φ m (s )σˆ θact (s ) − σθact(s )

2.3.5. Mixture-level constitutive relations

2.4. Equivalent thin-walled artery: Stability analysis

2.4.1. Nonlinear, non-autonomous system

ρ˙ R = k(ϒ − 1 )(ρR − ρoe ) (86)

σˆ θact = λ2θ (σθ , ρR )Sˆθact (σθ , ρR ) . (106)

2.4.2. Mechanobiological equilibrium

2.4.3. Mechanobiological stability

ς h − ς o <  ⇒ yh − yo < δ. (119)

df(yh (ζh ), ζh ) ∂ f(yh , ζh ) dyh (ζh ) ∂ f(yh , ζh )

which is valid in a neighborhood of the evolved homeostatic solution y = yh .

l1o = −ko (129)

2.5. Illustrative results for the murine aorta

3.1. Mechanobiological stability of equilibrium solutions

3.1.1. Acute increase in pressure

3.1.2. Elastin degradation

3.2. Mechanobiological stability of dynamic evolutions

3.2.1. Gain parameters Kσ and Kτ

3.2.2. Rate parameter ko

3.2.3. Material-to-prestress stiffness ratio cθ θ /(2σ θ )

3.2.4. Active contraction of muscle

Artery mass density ρ 1050 kg/m3

The authors declare that they have no conﬂict of interest.

Appendix A. Constitutive tangent moduli

Mechanobiologically equilibrated values

σθeo = φoe σˆ θe = φoe Sˆθe Geθ2 , σˆ zoe = Sˆze Gez2 (A.8)

Appendix B. Jacobian matrix derivatives

∂ϒ ∂ (τw /τwo ) o 3Kτ K̄τ

∂ϒ ∂ ((σθ + σz )/σIo ) ∂ (τw /τwo ) o Kσ 3Kτ K̄σ + K̄τ

∂ϒ ∂ ((σθ + σz )/σIo ) o Kσ K̄σ

Jacobian derivatives at original homeostatic state o

You might also like

ς h − ς o < ⇒ yh − yo < δ. (119)